0 Up votes0 Down votes

3 views59 pagesHow to invest in illiquid assets

Feb 13, 2015

© © All Rights Reserved

PDF, TXT or read online from Scribd

How to invest in illiquid assets

© All Rights Reserved

3 views

How to invest in illiquid assets

© All Rights Reserved

- Resume
- Glosssry-IVCA
- 2010 Preqin Infrastructure Review Sample Pages
- CEO/Busines Leader
- Top 50 Real Estate PE Firms
- Nyabwanga et al 2013
- HDFC Short Term Fund Apr11
- Water Infrastructure
- liquidity2081.pdf
- Directory of Indian Law Firms
- Demystifying the Financial Sector
- 1205-HKVCA VC Brown Paper
- HEALTH INSURANCE Healthcare Industry Structure_US Congress
- Camels Mels
- Liquidity
- Project Report on NPA
- BF - final
- 1864446
- A Demon of Our Own Design Book Notes
- Shajalal Islami Bank Ar 2013

You are on page 1of 59

Adrian Buss

Raman Uppal

Grigory Vilkov

Abstract

Alternative assets, such as private equity, hedge funds, and real assets, are

illiquid and opaque, and thus pose a challenge to traditional models of asset allocation. In this paper, we study asset allocation and asset pricing in a generalequilibrium model with liquid assets and an alternative risky asset, which is

opaque and incurs transaction costs, and investors who differ in their experience

in assessing the alternative asset. We find that the optimal asset-allocation strategy of the relatively inexperienced investors is to initially tilt their portfolio away

from the alternative asset and to hold more of it with experience. Counterintuitively, a decrease in the transaction cost for the alternative asset increases the

portfolio tilt at the initial date, and hence, the liquidity discount. Transaction

costs may induce inexperienced investors to hold a majority of the illiquid asset

at later dates, even if they are pessimistic about future payoffs, and produce a

sizable liquidity discount. During periods when the alternative asset is illiquid,

investors trade the liquid equity index instead, leading to strong spillover effects.

Keywords: portfolio choice, alternative assets, private equity, transaction costs,

heterogeneous beliefs, incomplete markets.

JEL: G11, G12

We thank Bernard Dumas for extensive discussions. We are grateful for comments from N

oel Amenc,

Harjoat Bhamra, Victor DeMiguel, Vihang Errunza, Denis Gorea, Rob Heinkel, Christian Heyerdahl-Larsen,

Francis Longstaff, Lionel Martellini, Ludovic Phalippou, Sergei Sarkissian, Eli Talmor, Nikolas Tessaromatis,

Josef Zechner, and seminar participants at Goethe University Frankfurt. Adrian Buss is affiliated with

INSEAD, Boulevard de Constance, 77305 Fontainebleau, France; Email: adrian.buss@insead.edu. Raman

Uppal is affiliated with CEPR and Edhec Business School, 10 Fleet Place, Ludgate, London, United Kingdom

EC4M 7RB; Email: raman.uppal@edhec.edu. Grigory Vilkov is affiliated with SAFE and Frankfurt School of

Finance & Management, Sonnemannstrasse 9-11, D-60314 Frankfurt, Germany; Email: vilkov@vilkov.net.

Introduction

Traditional models of portfolio choice and asset pricing assume that assets are liquid and can

be traded without cost. While this is a reasonable assumption for government securities and

publicly traded stocks of large companies, there are substantial costs for trading alternative

asset classes, such as private equity, stocks of smaller companies, hedge funds, and real-estate

funds,1 and, consequently, the interval between trades in these assets can span long periods.2

Moreover, because these alternative assets do not have long histories and regularly observed

market values, the returns from investing in these assets are less transparent than the returns

from public equity.3 As Malmendier and Nagel (2011, 2014)) show, investors expectations

are influenced by their personal experience even in transparent markets. Accordingly, we

allow investors experience to play a crucial role in forming expectations for alternative asset

classes, given the limited length and quality of their return data.4

Given the large and increasing role of alternative assets in portfolios,5 our objective is

to develop a model of asset allocation and asset pricing that takes into account both key

characteristics of alternative assets: (1) trading is costly, and (2) returns are opaque, and

thus, investors assessments of these assets depend on their experience. Specifically, when

assets can be traded only at a substantial cost, asset-allocation decisions depend not just on

current market conditions and investors experience, but also on past experience and past

asset-allocation decisions, in addition to expected future experience and trading costs.

Thus, the interaction between opaqueness and illiquidity raises fundamental questions

about asset allocation, asset pricing and their dynamics. How does the interaction between

opaqueness and illiquidity affect an investors portfolio? Should inexperienced investors hold

1

Collett, Lizieri, and Ward (2003) estimate transaction costs of over 3% for institutional real estate,

Beber and Pagano (2013) find that bid-ask spreads for smaller firms can be 10%, and costs of private equity

transactions can exceed even 10% (Preqin Special Report (2013a)).

2

Ang (2013) finds that even within liquid asset classes, subclasses can be highly illiquid, e.g., stocks in

pink-sheet OTC markets may not trade for a week, and municipal bonds typically trade twice per year.

3

The difficulties encountered in estimating returns of opaque assets are discussed by Phalippou (2009),

Phalippou and Gottschalg (2009), and Ang and Sorensen (2013) for private equity, by Dhar and Goetzman

(2005) for real estate, and by Ang, Ayala, and Goetzman (2014) for hedge funds and private equity.

4

The empirical findings regarding the importance of personal experience are mirrored by the following

quote from Albert Einstein: The only source of knowledge is experience.

5

Goetzman and Oster (2012) and Cejnek, Franz, Randl, and Stoughton (2014) report dramatic shifts

toward alternative assets for endowment funds. While in 1985 Yales asset allocation for alternative assets

was 20%, in 2013 it invested 79.8% in alternatives, mostly absolute return, private equity, and real assets.

alternative assets at all, and how should they revise their portfolios over time as they gain

experience but face substantial transaction costs? What risks arise because of illiquidity

and investors inexperience? What are the consequences for asset prices and risk premia

and their evolution over time? Finally, what are the dynamics of the optimal risk-sharing

arrangement between experienced and inexperienced investors?

To answer these questions, we develop a discrete-time model of a general-equilibrium

exchange economy with heterogeneous investors who have Epstein and Zin (1989) and Weil

(1990) utility functions. Our framework has a risk-free bond and two risky assets that differ

in their level of liquidity and opaqueness. We interpret the first risky asset to be a liquid

asset, such as the S&P 500 equity index that can be traded without cost, and the second

asset to be a relatively illiquid and opaque alternative asset, which incurs a proportional

transaction cost when traded. A transaction cost of less than 100% gives rise to endogenous

illiquidity; that is, periods where investors choose not to trade the alternative asset. The

extreme case of 100% transaction cost corresponds to the blackout studied in Longstaff

(2009), which captures exogenous liquidity, such as lockout periods imposed by hedge funds.

In addition to transaction costs, a second feature of our model is that it has two classes

of investors who differ in their experience, captured through differences in the precision of

their prior beliefs about the expected growth rate of the alternative assets dividends. In

order to highlight the effect of experience, one investor is assumed to have an infinitely

precise estimate of the expected growth rate of the alternative assets dividends, while the

other investors estimate is much less precise.6 Consequently, compared to the experienced

investor, while using Bayes law to update her beliefs, the inexperienced investor assigns more

weight to recent data and less weight to her prior. So, even if investors start out with beliefs

that have the same prior mean, after observing the dividends of the alternative asset, their

posterior beliefs differ with respect to both the posterior mean and the confidence (posterior

variance) they have about these beliefs. This heterogeneity in posterior beliefs drives the

dynamics of trade and asset prices in our model.

6

We have also analyzed the setup where both investors have imperfect precision, with one investors

precision being higher than that of the other. However, the results are very similar, so we focus on the

simpler case in which only one investor has imperfect precision.

There is strong empirical evidence that experience has a substantial effect on financial

decision making. Vissing-Jorgenson (2003) shows that the beliefs of retail investors depend

on their return experience, with inexperienced investors having the highest stock market

expectation at the peak. Greenwood and Nagel (2009) show that this result holds even for

well-trained, professional fund managers. Malmendier and Nagel (2011, 2014) show that

investors experience of macroeconomic shocks affects their beliefs and risk-taking.7

Our key results are summarized below. When one asset is opaque, the inexperienced

investor recognizes the estimation risk inherent in this asset, and so she reduces her initial

holdings of this asset. This is accompanied by an increase in the holding of the liquid equity

index, and, driven by the desire for precautionary savings, a substantial increase in holdings

of the bond. On the other hand, because of market clearing, the experienced investors initial

holdings of the alternative asset increase with its opacity. Consequently, neither investor

holds the fully diversified market portfolio. As investors gain experience, they increase

their investment in alternative assets. These asset-allocation characteristics are consistent

with the empirically observed behavior of institutional investors: Lerner, Schoar, and Wang

(2008) and Goetzman and Oster (2012) find that private (Ivy League) endowment funds

have increased their target allocation for alternative assets over the last decades as they

acquired (positive) experience and, nowadays, strongly overweight alternative assets in their

portfolios.8 In contrast, they find that relatively inexperienced endowment funds invest only

marginally in alternative assets. A consequence of this asset allocation in our model is that

the volatility of the inexperienced investors consumption growth is lower than that of the

experienced investor. However, in exchange for insuring the inexperienced investor, the

experienced investor earns a higher expected return on her wealth. In line with this, Lerner,

Schoar, and Wang (2008) find that endowment funds that entered earlier into alternative

assets achieved higher returns and attribute this success to the importance of experience in

interpreting ambiguous data.

7

Additional evidence of learning by trading can be found in Kaustia and Kn

upfer (2008), Seru,

Shumway, and Stoffman (2010), and Barberis, Greenwood, Jin, and Shleifer (2014). There exist also a number of papers using experiments whose results support our modeling of experience, e.g., Smith, Suchanek,

and Williams (1988), Dufwenberg, Lindqvist, and Moore (2005), Haruvy, Lahav, and Noussair (2007), and

Hussam, Porter, and Smith (2008).

8

See footnote 5 for evidence on the overweighting of alternative assets.

Surprisingly, because of the interaction between inexperience and illiquidity, the inexperienced investors optimal initial holdings of the alternative asset can be larger with

illiquidity. The intuition for this is that ideally the investor would like to start with a low

holding of the alternative asset and increase this position over time as she gains experience. However, this would generate substantial trading costs. Accordingly, inexperienced

investors that are new to an alternative asset class trade off the higher risk of initial overinvestment versus the cost of future rebalancing. If the transaction cost dominates, it is

optimal for such investors to reduce the portfolio tilt away from the alternative asset, and

thus, hold more initially. Transaction costs also induce portfolio inertia. Consequently, in

the presence of transaction costs, the inexperienced investor could end up holding a majority of the alternative asset over time even if she is more pessimistic about its growth rate

than the experienced investor. This effect is driven by prior beliefs and the reluctance to

change portfolio holdings in the presence of transaction costs.

As expected, transaction costs for trading the alternative asset reduce its turnover.

This also triggers a decrease in bond turnover because the bond is used to finance trade in

the alternative asset. However, the turnover of the liquid risky asset increases because it

is used as a substitute for trading the alternative asset. This spillover effect can be so

significant that the inexperienced investors holdings in the liquid risky asset track perfectly

the dividend dynamics of the alternative asset, introducing excess correlation and leading

to more volatile holdings in the liquid equity index.

With transaction costs for the alternative asset, its price drops and this is mirrored by

an increase in its expected return and also its volatility. For a transaction cost of 5%, we

find an average price discount of more than 4% and a maximum discount of more than 12%,

consistent with empirical data. For example, Franzoni, Nowak, and Phalippou (2012) find

an illiquidity discount of 10% for private equity and Hege and Nuti (2011) report valuation

gaps of 5%10% in private-equity secondary markets. The price discount also implies a

sizable increase in expected returns, in line with empirical evidence. Contrary to intuition,

the illiquidity discount is stronger for moderate transaction costs than for full illiquidity.

That is, the price discount is larger in a setting where investors incur a cost for trading the

asset, compared to a setting where trade is precluded exogenously. Because the bond and

the equity index serve as substitutes to the alternative asset, their prices increase in the

transaction costs for the alternative asset.

In addition to the economic insights described above, our paper also makes a technical

contribution: we demonstrate how to identify the equilibrium in a recursive fashion in a

model where markets are incomplete, the decision to trade or not is endogenous, the riskfree interest rate is endogenous, investors have heterogeneous beliefs that depend on their

experience, and preferences are given by Epstein-Zin-Weil utility functions.

The rest of the paper is organized as follows. In Section 2, we discuss the existing

literature that is related to our work. In Section 3, we describe the general model of the

economy we study. In Section 4, we characterize the equilibrium in this economy and explain

how it can be described by a system of path-independent backward-only equations, instead

of a system of backward-forward equations. In Section 5, we analyze the effect of transaction

costs and differences in experience on asset allocation, asset prices, and consumption. We

conclude in Section 6. Technical results are relegated to the appendix.

Related Literature

The alternative asset that we study has two key features: one, its return is opaque, so

experienced and inexperienced investors typically disagree about the expected growth rate

of its dividends; and two, trading the alternative asset is costly. Thus, our work is related

to two streams of the literature, each of which considers one of the features in isolation.

Two important recent papers provide a comprehensive analysis of the implications

of experiential learning in overlapping generation models assuming that young investors

disregard history and learn only from their own experiences. While investors in CollinDufresne, Johannes, and Lochstoer (2014) have recursive preferences, Ehling, Graniero,

and Heyerdahl-Larsen (2014) obtain an elegant analytical solution for log investors. Earlier

papers studying the effects on asset prices of differences in beliefs, but without experiential learning, include Harrison and Kreps (1978), Detemple and Murthy (1994), Zapatero

(1998), Scheinkman and Xiong (2003), Dumas, Kurshev, and Uppal (2009), and Bhamra

and Uppal (2014). See Basak (2005) for an extensive survey of asset-pricing models with

disagreement between investors and Morris (1995) for a philosophical discussion of models

where agents have difference in beliefs but do not learn from each others behavior; that is,

they agree to disagree.

Second, there are several papers that study the effect of transaction costs and illiquidity on portfolio selection and equilibrium asset prices. In an influential paper with many

insights, Longstaff (2009) shows that the effect of illiquidity on portfolio choice and asset

prices can be substantial. He considers a setup with two log-investors who differ only in

their rates of subjective time preference and face an exogenously specified blackout period, in which the asset cannot be traded at all. In an important extension to the work

by Longstaff (2009), Ang, Papanikolaou, and Westerfield (2014) study asset allocation in

a partial-equilibrium model where the duration and frequency of the blackout period are

stochastic. Vayanos (1998) and Acharya and Pedersen (2005) consider illiquidity in the

form of transaction costs in models where the investors trading decisions are exogenously

specified, e.g., for life-cycle reasons. Heaton and Lucas (1996) consider a general equilibrium model, but with quadratic transaction costs. Buss and Dumas (2014) study an

equilibrium model with transaction costs, assuming that investors have power utility and

full knowledge about all the parameters in the model. Garleanu and Pedersen (2014) provide portfolio choice and asset-pricing implications in a very tractable framework based on

market-impact costs. For excellent surveys of the literature on portfolio choice and asset

pricing with illiquidity, see the review papers by Amihud, Mendelson, and Pedersen (2005)

and Vayanos and Wang (2011, 2012, 2013).

The main distinguishing features of our model from existing work are the following:

we consider both experiential learning and illiquidity, which is modeled via the presence

of transaction costs, so investors choose endogenously whether or not to trade; investors

have recursive preferences that capture wealth effects (in contrast to exponential utility),

intertemporal hedging demands (in contrast to log utility), and a preference for early or late

resolution of uncertainty (in contrast to power utility); and, an endogenous risk-free rate.

Moreover, in contrast to the early literature studying the effect on asset prices of differences

in beliefs, in our model investors differ with respect to both the posterior mean belief and

confidence (posterior variance) they have in this belief.

Model

In this section, we describe the general-equilibrium model we use for our analysis. The model

is set in discrete time with a finite horizon: t {1, ..., T }. We assume that there exists a

single consumption good. The economy is populated by two types of investors, indexed by

k {1, 2}, who derive utility from consumption and differ in their experience. Investors can

trade three assets, indexed by n {0, 1, 2}, differing in their degree of riskiness, liquidity,

and opacity. Asset n = 0 is liquid and risk-free; asset n = 1 is liquid but risky; and asset

n = 2 is risky but illiquid with a dividend process that is opaque. In the rest of this section,

we give the details of these financial assets and investors.

3.1

We consider a Lucas (1978) economy with two trees, n {1, 2}, each generating dividends d(n, t). We assume that both dividends follow a binomial process, mimicking a

D

Brownian motion with drift D

n and volatility n , implemented as in Jarrow and Rudd

(1983). That is, an up-move (down-move) in the tree implies a relative change in dividend

1

D 2 D with unconditional probabilities of 1/2 each. To capture

of exp D

n

n

n

2

a correlation between the dividend processes, we set the conditional probability of an

up-move (down-move) in the first dividend tree, conditional on an up-move (down-move)

in the second dividend tree, to be = (1 + ) /2.

There exist two risky assets, each modeled as a claim to the dividends of one of the

trees. We normalize the number of shares outstanding for each asset to one. In addition to

these two risky assets, there exists a one-period discount bond in zero net supply, indexed

by n = 0. The number of shares of a particular asset n held by investor k at date t is

denoted by (n, k, t); and the price of each risky asset n as perceived by investor k at date

t, S(n, k, t), is determined in equilibrium with the price on the terminal date being zero.

3.1.1

Transaction Costs

We assume that the discount bond and the first risky asset are perfectly liquid; that is, can

be traded at any time without incurring costs. Thus, we interpret the bond as a risk-free

government bond and the first risky asset as an equity index.

In contrast, investors pay a proportional cost for trading the second risky asset, labeled

the alternative asset. The transaction cost () depends on the (dollar) value of the trade:9

(2, k, t), (2, k, t 1), S(2, k, t) (2, k, t) (2, k, t 1) S(2, k, t) ,

(1)

where (2, k, t) denotes the number of shares of the alternative asset held by investor k at

date t, and is the proportional transaction cost.10 If equals zero, the asset can be traded

without costs. On the other hand, if = 100%, the asset becomes fully illiquid, resulting in

a blackout period over which the asset cannot be traded at all, as in Longstaff (2009). To

minimize the effect of the initial holdings on the equilibrium outcome, we assume that the

investors can trade the alternative asset at no cost at the initial date. Thus, the optimal

initial holdings depend only on the initial wealth of the investorsnot its composition.

Moreover, in the case of a full blackout, it gives the investors a chance to adjust their

holdings once before the market shuts down, which allows us to compare our results with

those in Longstaff (2009).11

9

We could also consider the case where the transaction costs depend on the number of shares being traded,

as in Vayanos (1998). Transaction costs could differ also across investors and over time. One could also

allow them to vary with the state of the economy, for instance, being higher when aggregate consumption

is low. However, for ease of exposition, we focus on the case of constant and identical costs.

10

Because we are in a general-equilibrium setting, rather than assuming that the transaction cost is a

deadweight loss to society, we assume that the transaction cost is added back to the investors consumption

after the investor has made her consumption and portfolio decisions; this eliminates any wealth effects arising

from transaction costs.

11

Our model easily allows for transaction costs at the initial date. In the quantitative analysis of our

model, we study this case as a robustness exercise.

3.2

Given that we wish to study how differences in experience between investors and illiquidity

impacts asset allocation and asset pricing, we assume that each investor has the same initial

endowment, half a share of each of the risky assets and a zero position in the bond, and

homogeneous preferences.12 Specifically, investors are assumed to have preferences that are

of the Kreps and Porteus (1978) type. We adopt the Epstein and Zin (1989) and Weil (1990)

specification of this utility function, in which lifetime utility V (k, t) is defined recursively:

h

1

1

V (k, t) = (1 ) c(k, t)

Etk

1 i

1 1

V (k, t + 1)

,

where Etk denotes the conditional expectation at time t under the investors subjective

probability measure, c(k, t) > 0 is the consumption of investor k at date t in state (t, s),13

denotes the subjective rate of time preference, > 0 is the coefficient of relative risk

aversion, > 0 is the elasticity of intertemporal substitution, and =

1

11/ .

The Epstein

and Zin (1989) and Weil (1990) utility specification nests the more standard time-separable

utility functions, but has the well-known advantage that the risk aversion parameter, which

drives the desire to smooth consumption across states of nature, is distinct from the elasticity

of intertemporal substitution parameter, which drives the desire to smooth consumption

over time. The importance of using these utility functions in a model with learning is

highlighted in Collin-Dufresne, Johannes, and Lochstoer (2013).

We now explain how we model the role of experience. We assume that both investors

have perfect knowledge with respect to the dividend dynamics of the equity index and also

know the correlation between the two dividend trees. For the alternative asset, which is

subject to transaction costs, we assume that the first investor (the experienced investor)

has perfect knowledge about its dividend dynamics; that is, the investors beliefs coincide

with the objective probability measure. But, the second investor (the inexperienced investor) has imperfect information about the assets dividend dynamics and updates her

12

We could easily allow the investors to differ along all three dimensions of their utility functions: time

preference, risk aversion, and elasticity of intertemporal substitution. However, our focus is on the effect of

illiquidity and investors experience rather than preference heterogeneity.

13

To simplify notation, we do not write explicitly the dependence on the state (t, s).

beliefs after each realization of the alternative assets dividends using Bayesian learning.

Thus, illiquidity and opacity reside jointly in the alternative asset.

Specifically, given the prior information and her observations, the inexperienced investor

has to infer the probability for an up-move, p2,u , in the binomial tree of the alternative asset.

A conjugate prior for p2,u is the Beta distribution, Beta(1 ; 2 ), where the hyperparameters

1 and 2 can be interpreted as virtual (pseudo) counts of up- and down-moves before the

start of the economy (prior knowledge). While larger values for 1 and 2 are in general

associated with more precise prior knowledge about the distribution, the relative share of

1 versus 2 captures the prior expected value for p2,u , as outlined below. Starting with

the Beta prior and having observed u up-moves and d down-moves in the alternative assets

binomial tree, the investors posterior beliefs are Beta distributed, with p2,u Beta(1 +

u, 2 +d).14 This implies that the posterior mean probability under the investors subjective

probability measure for the next move in the alternative assets tree being an up-move is

Et [p2,u ] =

1 + u

,

1 + u + 2 + d

(2)

Var(p2,u ) =

(1 + u)(2 + d)

.

(1 + u + 2 + d)2 (1 + u + 2 + d + 1)

(3)

That is, the inexperienced investor updates her beliefs by simply counting the number

of observed up-moves u and down-moves d in the tree. The predictive expected probability of the next move being up (down) is then just the relative share of observed upmoves (down-moves), incorporating in addition, the prior information corresponding to

the pseudo-counts. To incorporate the assumption that the inexperienced investor knows

the correlation between the dividends for the two assets, we set the subjective conditional

probability of an up move (down move) equal to the objective conditional probability.

Our learning algorithm highlights the importance of learning from experience. Specifically, the inexperienced investor acts as a trend-chaser, revising her beliefs upwards after

a positive shock and downwards after a negative shock. There is substantial evidence in the

14

10

literature for this type of learning behavior. For example, Malmendier and Nagel (2011,

2014) show that investors experience of macroeconomic shocks affects their beliefs and risktaking. Using data from the tech-bubble in the late 1990s, Vissing-Jorgenson (2003) shows

that the beliefs of retail investors depend on their return experience, with inexperienced

investors having the highest stock market expectation at the peak, and Greenwood and

Nagel (2009) show that this result holds even for well-trained, professional fund managers.

In this section, we first describe the optimization problem of each investor. We then impose

market clearing and explain how one can solve for the equilibrium.

4.1

The objective of each investor is to maximize expected lifetime utility by choosing consumption, c(k, t), and the portfolio positions in each of the financial assets, (n, k, t):

V (k, t) =

max

{c(k,t),(n,k,t)}

1

1

(1 ) c(k, t)

Etk

1 i

1 1

,

V (k, t + 1)

(4)

c(k, t) +

2

X

(n, k, t)S(n, k, t) + (2, k, t), (2, k, t 1), S(2, k, t)

(5)

n=0

(0, k, t 1) +

2

X

(n, k, t 1) S(n, k, t) + d(n, t) .

n=1

constraint (2, k, t) 0.

The left-hand side of budget equation (5) is the amount of wealth allocated to consumption and the purchase of assets at date t, including transaction costs for trading the

alternative asset, as specified in equation (1). The right-hand side captures the unit payout

of the short-term bond as well as the prices and dividends of the two long-term risky assets

11

at date t, scaled by the number of shares of each risky asset purchased at date t 1. This

sum can be interpreted as the wealth with which the investor arrives at date t.

After some simplifications and rearranging, one gets the following kernel conditions:15

S(0, k, t) =

Etk

S(1, k, t) =

Etk

S(2, k, t) =

M (k, t + 1)

,

M (k, t)

(6)

M (k, t + 1)

S(1, k, t + 1) + d(1, t + 1) ,

M (k, t)

"

1

k M (k, t + 1)

E

1 SSC (k, t) t

M (k, t)

(7)

S(2, k, t + 1) + d(2, t + 1)

(2, k, t)

!#

(2, k, t), (2, k, t 1), S(2, k, t)

1

1 SSC (k, t)

(2, k, t)

(8)

where SSC is associated with the Lagrange multiplier on the shortsale constraint and

M (k,t+1)

M (k,t)

1

M (k, t + 1)

=

M (k, t)

1)1

V (k, t +

k

Et [V (k, t + 1)1 ]

1

c(k, t + 1)

c(k, t)

1

Note, the kernel condition for the alternative asset (8), equating the benefits from holding

the asset to the benefit from selling the asset, deviates in two dimensions from the traditional kernel condition for fully liquid assets, as given in equation (7). First, the benefits

derived from holding as well as selling the asset are both net of transaction costs, i.e., they

are adjusted for costs associated with an immediate sale, through (2, k, t), (2, k, t

1), S(2, k, t) /(2, k, t), as well as for potential future costs in case of a sale in the next pe

riod, given by (2, k, t + 1), (2, k, t), S(2, k, t + 1) /(2, k, t). Second, the pricing kernel

contains the shadow cost stemming from the shortsale constraint, reflected by

15

12

1

1SSC (k,t) .

4.2

Market-Clearing Conditions

In the economy we are considering, markets exist for the one-period risk-free bond, the two

risky securities, and the consumption good. The market-clearing conditions for the financial

assets imply that aggregate supply must equal aggregate demand:

ss (n) =

2

X

(9)

k=1

where ss (n) denotes the aggregate supply of asset n. For the two risky assets, aggregate

supply ss (n), n {1, 2} is equal to 1, and for the risk-free bond, we have that ss (0) = 0.

If the financial markets clear, then Walras law ensures that aggregate dividends must

equal aggregate consumption:

2

X

d(n, t) =

n=1

4.3

2

X

c(k, t).

k=1

Definition of Equilibrium

Equilibrium in this economy is defined as a set of consumption policies, c(k, t), and assetallocation policies, (n, k, t), along with the resulting price processes for the financial assets,

S(n, k, t), such that the consumption policy of each investor maximizes her lifetime utility,

that this consumption policy is financed by the optimal asset-allocation policy, and markets

for financial assets and the consumption good clear.

The budget constraint (5), the kernel conditions (6), (7), and (8), together with the complementary slackness conditions associated with the shortsale constraint, and the marketclearing conditions in (9) form the system of equations that characterize the equilibrium.

4.4

If financial markets are complete, one can separate the task of identifying the equilibrium

into two distinct steps by exploiting the condition that investors can achieve perfect risk

sharing. This condition can be used to first identify the optimal allocation of aggregate

13

consumption across investors (central planners problem). Then, in the second step, using

these consumption choices one can determine asset prices and also the portfolio policy of

each investor that supports this allocation.

In contrast, when financial markets are incomplete, one cannot separate the task of

identifying the equilibrium into two steps because the consumption allocation one chooses

must lie in the span of traded assets. Instead, one must solve for the consumption and

portfolio policies simultaneously. In principle, one can identify the equilibrium by solving

simultaneously the set of nonlinear first-order conditions for the investors along with the

market-clearing conditions for all the states across all dates, as proposed in Cuoco and He

(2001). The problem in implementing this approach is that the number of equations grows

exponentially with the number of periods, so that a recursive approach would be preferable.

However, there are several problems in solving this system of equations recursively in a

general-equilibrium setting. The first problem is that the current consumption and portfolio

choices depend on the prices of assets, which, as shown in Equations (6), (7) and (8), depend

on future consumption. But, in a general-equilibrium setting, when the investor attempts

to solve for the optimal consumption and portfolio policies at date t, asset prices need to

adjust in order for markets to clear. However, if one were solving the system of equations

backward, these prices could not adjust because they depend on future consumption, which

has already been determined in the previous step. Thus, to solve these equations, one

would need to iterate backward and forward until the equations for all the nodes on the

tree are satisfied. Dumas and Lyasoff (2012) address this problem by proposing a timeshift whereby at date t one solves for the optimal portfolio for date t but the optimal

consumption for date t + 1, instead of the optimal consumption for date t. Using this

insight allows one to write the system of equations so that it is recursive.

Additional difficulties arise because of transaction costs, because whether a particular

security is traded or not at a given node is now determined endogenously. Specifically, if

investors choose to trade each of the assets, then they will agree on the prices of these

assets. However, if investors find it optimal not to trade some of the assets, then investors

will disagree on the prices of the assets that are not traded at that node. Consequently, the

14

system of equations characterizing the solution depends on whether or not investors choose

to trade all assets or only some of the assets.16

To address these problems, we observe that the past portfolio holdings enter the system

of equations only through condition (8), as a first partial derivative of the transaction cost

function () with respect to the current portfolio investment. Under the assumption that

the transaction costs are a constant proportion, , of the value of the asset being traded,

there are only three possibilities for the form of this derivative. It is equal to zero when an

investor decides not to trade; it is equal to S(2, k, t) when the investor decides to increase

the position in the asset; or, it is equal to S(2, k, t) when the investor sells the asset.

Consequently, all the (2, k, t 1) values for which the investor decides to buy (sell) an asset

at time t result in the same solution (2, k, t) for a given value of current consumption c(k, t).

In other words, instead of solving the problem over the undetermined wide grid of portfolio

holdings at t 1, we can solve it first for the two trading decisionssell or buyat time t.

The solution to this provides us with the bounds of the no-trade regionthe region for which

the portfolio investment from t1 to t does not change. Within these bounds of the no-trade

region that we have determined we use the past portfolio holdings as an endogenous state

variable. We then solve the system of equations explicitly by restricting current portfolio

holdings to equal past portfolio holdings and, in turn, eliminate the condition that investors

agree on asset prices.

Using the time-shift and the insights described above, one can solve for the equilibrium recursively. Finally, after solving the dynamic program recursively up to the initial

date, we undertake a simple single forward step for each simulated path of the underlying dividend processes to determine the equilibrium quantities. Additional details of this

recursive solution method are given in Appendix B.

16

For a discussion of the challenges in solving dynamic optimization problems with occasionally binding

constraints, see Christiano and Fisher (2000).

15

In this section, we wish to evaluate the impact of illiquidity and opaqueness on asset allocation and asset prices. To do this, we undertake a quantitative analysis of the model

described in the previous section.

5.1

For our numerical illustration, we focus on the U.S. market. Thus, we interpret the risk-free

asset as a U.S. Treasury bill and the risky, liquid asset as a broad U.S. equity index, such

as the S&P 500. Given the tight link between dividend and return volatility in an exchange

economy, we set the dividend volatility for the equity index equal to 20%, comparable to the

21.6% volatility reported in Longstaff (2009). We use a 1.2% expected growth rate for the

index dividends, based on the value reported in Beeler and Campbell (2012, their Table 2).

All returns are measured and compounded on a per annum basis.

In general, the alternative asset could represent any asset class that faces non-trivial

costs for trading and has only a short time-series of historical price data, e.g., private equity,

hedge funds, or real-estate funds. However, to give a better feel for the quantitative results,

we first match the parameters to the private-equity asset class, varying selected parameters

afterwards to accommodate other asset classes. Specifically, we interpret investments in the

illiquid asset as investments into several private equity funds with different vintage years,

e.g., for diversification purposes.17 We mainly study a horizon of T = 15 years, longer than

the typical ten years for single partnership contracts (Phalippou and Gottschalg (2009) and

Harris, Jenkinson, and Kaplan (2014)). Later we report results for also 10 and 20 years.

When modeling the payout process for private equity, one has to answer two important

questions: (1) Are its payouts similar to dividends? and (2) What are reasonable parameter

values for the mean and volatility of the payouts? First, based on our interpretation of

the illiquid investment being into several private-equity funds with different vintage years,

one can easily imagine a dividend-like payout structure. That is, while newly initiated

17

For example, as of June 30, 2014, Calpers reported 295 active private equity investments with vintage

years ranging from 1991 to 2014 and a median vintage year of 2006.

16

funds might still raise capital, the group of middle-aged and older funds would already

realize investments and distribute the proceeds, so that the aggregate cash flow resembles

a regular flow of dividends.18 Second, a broad range of empirical evidence suggests that

private-equity investments are more similar to traded securities than one might expect. For

example, Moskowitz and Vissing-Jorgensen (2002) find that a portfolio of all private equity

has a mean and standard deviation of returns close to that of the value-weighted index

of traded stocks. Similarly, Kaplan and Schoar (2005), Cochrane (2005), and Phalippou

and Gottschalg (2009) document comparable average returns for public and private equity,

and the evidence in Metrick and Yasuda (2010) supports a volatility comparable to that of

public equity.19 In light of this evidence, for our base case we chose the same mean and

standard deviation for private equity as for the equity index. This also allows us to isolate

cleanly the effects of illiquidity and opaqueness from potential differences in the underlying

payout process.

We set the correlation between dividends of public and private equity to be 50%, relying on Longstaff (2009), which is consistent also with the 40%60% correlation between

the assets returns, documented in Ang, Papanikolaou, and Westerfield (2014) and Welch

(2014).20 Finally, we choose the share of initial income of the assets to match the ratio of

the market capitalization of the assets. Specifically, we set the share of income provided by

private equity to be 20% and that from the equity index to be 80%, so that the capitalized value of the alternative asset is consistent with the $3.5 trillion value of assets under

management for private equity relative to $18 trillion U.S. stock-market capitalization.21

Transactions in alternative assets entail different types of costs. For example, transactions in private equity include participation in the initial offering and purchases or sales in

the secondary market, so there would be costs for identifying the appropriate private eq18

Calpers describes the aggregation as follows: As the portfolio matures and distributions grow, the

[Private Equity] Program is becoming self-sustaining.

19

Metrick and Yasuda (2010) find a 60% volatility and 20% correlation for individual buyout firms and

report that funds on average invest into 15 firms, implying an individual fund volatility of about 25%. Given

that the investment into several funds yields additional diversification benefits, a volatility of around 1520%

for private equity as an asset class seems reasonable.

20

The returns have a higher correlation than the underlying cash flows, due to the pricing-kernel channel.

21

See Preqin Special Report (2014) for private-equity data; for stock-market capitalization, data are

provided by the World bank at http://data.worldbank.org/indicator/CM.MKT.LCAP.CD

17

uity funds, closing the transaction, hiring specialized advisers or, in the secondary market,

searching for a counterparty. In our model, we do not differentiate between these different

types of costs; instead, we consider proportional transaction costs up to 10%, as reported

by Preqin Special Report (2013a), as well as 100% to capture the case of full illiquidity.

In the second stage of our quantitative analysis, we vary selected parameters. To study

the impact of the investment horizon, we consider setups with 10 and 20 years. Next, we report results when the alternative assets dividend volatility is 12.5% or 27.5%, representing

the lower volatility of hedge funds and real estate or the higher volatility of venture capital

and small stocks.22 We also consider correlations of 0.25 and 0.75 with the former representing, for example, the low correlation between buyout funds and equity indices (estimated

to be 0.27 by Ang, Papanikolaou, and Westerfield (2014)) and the latter representing small

stocks that have a correlation of about 0.80 with the market. Similarly, we report results

for cases where the alternative assets share of income is 10% and 30%. The former captures

hedge funds and small stocks, and the latter represents commercial real estate.23

Our two groups of investors can be viewed as institutional investors, such as endowment

or pension funds, that both have access to alternative asset classes, but differ in their

experience, e.g., because of fund size or managers ability. Specifically, we assume that

the inexperienced investor has only little prior exposure to alternative assets, modeled as

a Beta(1, 1) prior distribution. Importantly, although having a lower precision, the prior

mean of the inexperienced investor on the initial date coincides with the true probability.

We also report results for different levels of prior precision.

Finally, we choose the preference parameters of the investors to match the risk-free rate,

equity risk premium and equity-market volatility. Our choices for relative risk aversion,

= 1.75, rate of time-preference, = 0.95, and elasticity of intertemporal substitution,

= 1.5, imply a risk-free rate of 1.1%, an equity market risk premium of 6.41%, and a

22

Buraschi, Kosowski, and Trojani (2014) report volatilities for hedge fund indices to range from 3.8%

to 16.0% and Ghysels, Plazzi, Torous, and Valkanov (2012) document commercial real estate volatilities to

range from 5.2% to 17.9%. In contrast, Ang, Papanikolaou, and Westerfield (2014) find a 27.8% volatility for

venture capital and data from Kenneth Frenchs web site implies a volatility of about 30% for small stocks.

23

Preqin Special Report (2013b) reports assets of $1.7 trillion for U.S. hedge funds while the S&P 600

market capitalization is below one trillion, implying a share of less than 10%. In contrast, Florance, Miller,

Peng, and Spivey (2010) report a value of $9.5 trillion for U.S. commercial real estate, which is about half

of U.S. stock market capitalization.

18

6.94%, and 20.17% estimated by Beeler and Campbell (2012). For comparative statics,

we also consider variations in the preference parameters. The base-cases values for the

parameters and the variations in them are summarized in Table 1.

5.2

The effects of experience on the inexperienced investors asset allocation exhibit themselves

through two forces. First, with each new dividend realization the inexperienced investor

updates her posterior mean for the probability of the next move in the alternative assets

dividend tree being an up-move. Second, with the passage of time, the inexperienced

investor updates also her posterior variance. One can interpret the posterior mean as

reflecting the investors optimism and the posterior variance reflecting her confidence.

Figure 1 illustrates the dynamics of the economy for two specific simulations. Panels A

and B show the evolution of the alternative assets dividend (left-hand-side vertical scale)

for each simulation, as well as the evolution of the posterior mean and posterior variance

of the inexperienced investors beliefs (right-hand-side vertical scale) about the probability

of an up-move in the alternative assets dividend tree.24 The shaded gray areas highlight

periods of negative dividend realizations; that is, periods in which the inexperienced investor

adjusts her posterior mean downwards.25 Note that the dividend dynamics for the equity

index coincide for both paths, as shown in Panel C.

The revision in the posterior mean for the alternative asset can be upwards or downwards, depending on whether the most recent dividend shock was positive or negative. For

example, in the first simulation (Panel A of Figure 1), the investor experiences a positive

shock at date 2, revising her expected growth rate upwards. However, the negative shock at

date 3 brings her back to her prior belief of 0.5. Finally, the inexperienced investor ends up

being optimistic because of the boom over the last 10 periods. In contrast, in the second

24

In the figure, we actually plot 2.5 times the posterior standard deviation in order to match the scale for

the posterior mean so that we can plot both quantities in the same figure.

25

For ease of exposition, we reproduce the shaded gray areas in all the figures that follow, facilitating their

interpretation.

19

simulated path (Panel B), the alternative assets dividends are first increasing, implying an

increase in the inexperienced investors optimism, before the dividends slowly decline over

time, rendering the investor pessimistic relative to the earlier period.

In contrast to the posterior mean, the posterior variance is generally declining over

time. But, the change is stochastic, depending not only on the passage of time, but also on

whether the shocks experienced were positive or negative. One can see this by comparing

across the two simulations the change in posterior variance between t = 2 and t = 3. In the

second simulation, the investor experiences a second successive positive shock, which leads

to a much stronger reduction in variance than for the first simulation, where the positive

shock at t = 2 is followed by a negative shock at t = 3. Note that in certain cases the

posterior variance can increase over time, as one can see in the second simulation, in which

it increases over the period t = 3 to t = 4. This occurs because at dates 2 and 3 the

alternative assets dividend shocks were positive but the shock at t = 4 is negative, which

makes the investor less confident about the beliefs she had formed on the basis of the

earlier positive shocks.

Below, we study the quantities of interest for two specific paths in the simulation and

also the average across 50,000 paths. Examining two particular paths allows us to illustrate

the impact of changes in the posterior mean and precision. Studying the average allows us

to understand whether the results for the two paths hold true in general, and also allows

us to focus on the impact of changes in the posterior variance, because the changes in the

posterior mean cancel out to a large extent when we average across paths.26

Note that for the experiential model of learning that we study, the posterior variance

decreases over time on average. For our choice of parameter values, the posterior standard

deviation decreases from 28.28% at the initial date to 12.25% at t = 15 (so, in the figure,

the decrease shown is 70.71% (= 2.5 28.28%) to 30.61% (2.5 12.25%).27 However, there

is still considerable variation in the posterior mean across sample paths, as illustrated in

26

Note that although the changes in the posterior mean cancel out when we average across paths, nonlinear

responses to the these changes will not cancel out perfectly.

27

Note that in a model with more than two states, the decline in the posterior variance with each observation would be even slower than it is in the model specified above.

20

Figure 2. For example, in more than 40% of the sample paths, the posterior mean differs

from the true mean by at least 0.10. That is, even though the investors confidence about

her posterior mean is very high after fifteen periods, she may be confident about a mean

that is substantially different from the true mean.

5.3

Asset Allocation

We start by investigating the implications of the model for the asset allocation choices of

the two investors. The optimal holdings (number of shares) of the inexperienced investor

for the three assets at the initial date are reported in the three panels of Table 2 and their

evolution over time is shown in Figures 35.28

If both investors were equally experienced, then it would be optimal for each investor to

hold 0.5 shares of each of the two risky assets (the market portfolio) and zero bonds. However, when one of the investors is relatively inexperienced, then neither investor holds the

market portfolio; instead, the inexperienced investor tilts her portfolio away from the alternative asset, reducing her holding of this asset from 0.5 to 0.123see Panel A of Table 2

and increasing her holdings of the bond and the equity index. The intuition for the portfolio

tilt away from the alternative asset is that the presence of estimation risk induces a negative

hedging demand for the alternative asset because of changes in perceived investment opportunities as the investor updates her beliefs. Because of the market-clearing condition, in

response to the inexperienced investors tilt away from the alternative asset, the experienced

investor tilts her portfolio toward the alternative asset and holds 0.877 shares. These results

for inexperienced investors are consistent with the results in partial-equilibrium models obtained by Brennan (1998), Barberis (2000), and Brandt, Goyal, Santa-Clara, and Stroud

(2005).29 In addition, the results for the experienced investor are consistent with the empir28

Because of market-clearing, the optimal holding of the more experienced investor for each of the risky

assets is one minus the holding for the inexperienced investor and for the bond it is the negative of the bond

holding of the inexperienced investor.

29

Williams (1977), Detemple (1986), Dothan and Feldman (1986), and Gennotte (1986) study the portfolio

problem of an investor who does not know the true state of the economy but who knows the stochastic process

governing the state variables. The key insight used in these papers to solve the investors problem is that

in continuous-time the investors problem can be separated into a filtering problem, in which the investor

estimates the current values of the state variables, and a standard investment problem in which the estimated

values of the state variables replace the state variables themselves. Brennan (1998) analyzes in continuous

time the simpler problem of an investor who is uncertain about the mean return on the risky asset. The

investor learns about the mean as she observes returns over time, which one can view as a special case of the

21

ically observed asset holdings of endowment funds, e.g., Lerner, Schoar, and Wang (2008)

and Goetzman and Oster (2012) find that large, experienced funds strongly overweight

alternative assets relative to the market and their peers.

The dynamics of the asset-allocation decision of the inexperienced investors holdings are

intriguing. Panel A of Figure 3 shows that on average the inexperienced investor increases

her holding of the alternative asset over time, with large increases in the first few periods.

This is because the largest decreases in the posterior variance come in the first few periods

(see Figure 1). We also find the surprising result that as the transaction costs for the

alternative asset increase, the inexperienced investor may hold a larger share of this asset.

For example, when the transaction cost increases from 2.5% to 10%, the initial holding of

the alternative asset increases from 0.041 shares to 0.144 shares, and when the transaction

cost is 100%, the holding of the alternative asset increases to 0.273 shares (see Panel A of

Table 2). The intuition for this is that ideally the inexperienced investor would like to build

up her position in the alternative asset slowly over time as she anticipates gaining experience

(posterior variance declining). However, this strategy entails substantial transaction costs.

Therefore, if future trading costs dominate, the inexperienced investor chooses to reduce

the portfolio tilt away from the alternative asset, and thus, holds a larger share of this

asset.30 This intertemporal tradeoff is evident from Panel A of Figure 3, which shows that

as the transaction costs increase, the initial holding increases while the investor cuts back

on future volume trading, which is reflected in the flattening out of the plot.

Panels B and C of Figure 3 illustrate the effects of changes in the posterior mean as the

inexperienced investor updates her beliefs. For example, in the case of the first simulated

path and low transaction costs (black solid line in Panel B), the inexperienced investor

substantially increases her holdings of the alternative asset due to the sizable change in her

posterior mean triggered by positive dividend news at t = 2. For the next few periods, she

analyses of Detemple (1986), Dothan and Feldman (1986), or Gennotte (1986), in which the dynamics of the

unobservable state variable, the mean return, are eliminated. Just like in the partial-equilibrium models of

Barberis (2000) and Brandt, Goyal, Santa-Clara, and Stroud (2005), in our general-equilibrium model the

investor learns about both the mean and the variance; in addition, investors also face transaction costs for

trading the alternative asset.

30

It is important to note that this result is not driven by the absence of transaction costs at the initial

date. If we had a transaction cost also on the initial date, the inexperienced investor would deviate even less

from her endowed holding of 0.5 shares, and so would be holding even more shares of the alternative asset.

22

does not change her holdings at all because the revisions in the posterior mean are too small

to compensate for the transaction cost. Only during the boom phase (dates 610), when the

posterior mean increases again, the investor strongly increases her holdingsin line with

her optimistic beliefs. In contrast, for transaction costs of 10% the inexperienced investor

never trades the alternative asset, because the changes in her beliefs are not substantial

enough to overcome the cost of trading. In the case of full illiquidity, trading is precluded

exogenously.

In the second simulated path, Panel C of Figure 3, the inexperienced investor increases

her holdings at the start of the economy as a reaction to the two positive dividend shocks at

dates 2 and 3. After this, the changes in dividends are small (alternating between negative

and positive shocks), so the investor does not rebalance her position in order to save on

transaction costs. However, in response to the series of negative shocks starting at t = 8,

indicated by the gray background, the inexperienced investor reduces her holdings, but only

for low levels of transaction costseven though her posterior mean is declining dramatically

over this period. This leads to the surprising result that the inexperienced investor ends

up holding a majority of the alternative asset, despite being pessimistic about it. Even

more surprising are the differences across the two simulations. For example, towards the

end of the economy and for transaction costs of 10%, in Panel C the inexperienced investor

holds more than twice as much of the alternative assetdespite being pessimisticthan in

Panel B, where she is highly optimistic. This contrast nicely showcases the crucial interplay

of experience and illiquidity and demonstrates the path dependence of asset holdings on the

evolution of the economy. This path dependence implies that when evaluating a portfolio,

it is not possible to say whether it is optimal or not unless one also knows the history of

the economic conditions under which this portfolio was formed.

We now turn our attention to the equity index. Panels B and C of Figure 4 illustrate that

the inexperienced investor uses the equity index, which is correlated with the alternative

asset, as a substitute when she updates her beliefs but trading the alternative asset is too

costly.31 For example, for the first simulated path (Panel B) and for transaction costs of 10%

31

Swensen (2009), the chief investment officer of Yale University, highlights this challenging interplay between illiquid assets and portfolio rebalancing, suggesting that choosing the least bad alternative, investors

23

or full illiquiditycases in which the investor abstains from trading the alternative asset

the holdings in the equity market track perfectly the dividend dynamics of the alternative

asset. That is, after a positive shock to the alternative assets dividends, the inexperienced

investor increases her investment in the liquid equity index, and vice versa. This spillover

or contagion effect materializes in an increase in the volatility of changes in the holdings

of the equity index by the inexperienced investorfor the case of 10% transaction costs we

find a relative increase in the volatility of the changes in the holdings of the equity index

of more than 14% and also excess comovement in returns.

Similarly, Panels B and C of Figure 5 illustrate that the bond holdings of the inexperienced investor are closely linked to the dividend dynamics of the alternative asset. That is,

after a positive dividend realization, the inexperienced investor is willing to borrow to trade

on this positive newseither by buying the alternative asset directly or using the equity

index as a substitute. The implications of an increase in precision can be clearly seen in

Panel A, which gives the average bond holdings over time. As the inexperienced investor

gains experience over time, she faces less estimation risk and, accordingly, reduces her bond

holdings as her precautionary-savings motive diminishes.32 The result that the change in

holdings of the alternative asset are offset by changes in bond holdings is consistent with

the empirical evidence in Brown and Tiu (2013).

We conclude this section by looking at the comparative statics for the asset-allocation

decision. Table 2 shows that the asset allocation implications for the variations from the

benchmark case are the same as for the base case. Strikingly, the portfolio tilt becomes

stronger with the investors investment horizon because of an increase in the size of the

negative hedging demand; that is, the magnitude of the hedging demand is increasing in

the investment horizon. For example, if the horizon increases from 15 to 20 years, the

should fund overallocations to private assets by reducing holdings of risky marketable securities to control

the portfolio risk.

32

Note, the average bond holdings of the inexperienced investor turn negative after some time. The reason

for this is that, because of the negative hedging demand for the alternative asset induced by estimation risk,

it is more likely that the inexperienced investor hits her shortsale constraint than that she owns all shares

of the alternative asset. Thus, she is less constrained in buying shares of the alternative assets, financed by

borrowing, than selling shares and investing the proceeds into the bond. This has a negative impact on the

inexperienced investors bond holdings which, when it dominates the precautionary savings motive, leads

to a short position in the bond. As shown in the figure, transaction costs dampen this effect because they

reduce the likelihood of the investor hitting her shortsale constraint.

24

inexperienced investor reduces her holding of the alternative asset from 0.123 shares to

0.022 shares in the absence of transaction costs. The magnitude of the portfolio tilt also

depends on the risk-return benefits delivered by the alternative asset. For example, for a

higher volatility of the alternative asset (e.g., small stocks or venture capital), and for a lower

dividend correlation (e.g., hedge funds or real estate), the inexperienced investor reduces

her tilt away from the market portfolio and, accordingly, holds more of the alternative

assets as the diversification benefits dominate the negative hedging demand stemming from

estimation risk. Similarly, when the relative size of the alternative asset in the economy

increases, the inexperienced investor reduces the tilt away from the market. This is driven by

the fact that a higher share of dividends increases the alternative assets expected return, as

shown by Cochrane, Longstaff, and Santa Clara (2008), so that the asset is more attractive

to hold and the effect of estimation risk becomes relatively less important. We also study

the dynamics of asset allocation for different levels of correlation, volatility, and precision

in Figure 6. While most dynamic effects are similar to the base case, a higher initial

precision implies smaller revision to her position over time and a lower correlation results

in considerably less spillover effect as the equity index is not as good a substitute.

We also study the comparative statics with respect to the preference parameters of the

investors; because these results are what one would expect, they are not reported in the

tables. For instance, as the investors risk aversion increases or her elasticity of intertemporal

substitution decreases, the inexperienced investor is less willing to tolerate the estimation

risk and associated variability in consumption over time, so that she reduces her holdings of

the alternative asset. In some cases, the investor even completely abandons diversification,

investing only in the bond and equity index.

5.4

Turnover

Table 3 shows that the effect of transaction costs on turnover is substantial.33 From Panel A

of Table 3, we see that when we introduce a 5% transaction cost for trading the alternative

33

We measure turnover as the (absolute) dollar volume traded, first averaged across all states on a given

date, and then across all dates.

25

asset its turnover drops to about half, and for a 10% transaction cost, the turnover drops by

two-thirds. The majority of the trading in the alternative asset is driven by the stochastic

nature of the inexperienced investors updating of her prior mean, as can be seen from

Panels B and C of Figure 3.34

From Panel B of Table 3, we see that the decrease in turnover of the alternative asset

is offset by an increase in turnover of the liquid equity index of up to 45% for our base case

in the presence of transaction costs of 10%. That is, trading the equity index becomes a

substitute for trading in the alternative asset, as illustrated in Panels B and C of Figure 4.35

In contrast, we observe from Panel C of Table 3 that the decrease in turnover of the

alternative asset is accompanied by a decline in the average turnover of the risk-free bond.

Finally, examining the results of the variation in the benchmark parameters, we can see

that the results are again qualitatively the same. Noticeably, the reduction in turnover is

stronger for the cases of a low dividend correlation, more precise priors of the inexperienced

investor, and a lower risk aversion (not reported). In all these cases, the investor has

already invested considerably more in the initial period, so that her need for future trading

in response to gains in experience are less crucial, implying that the trading costs dominate.

5.5

Asset Prices

When we move from the case where both investors have infinitely precise priors to the case

where one investor is inexperienced, there are two opposing forces that impact the price of

the alternative asset even in the absence of transaction costs. First, as explained by Dumas,

Kurshev, and Uppal (2009), sentiment risk arising from differences in beliefs reduces the

price of the asset. However, as explained by Miller (1977), shortsale constraints limit the

expression of negative sentiment, and so increase the price. In all our setups, the effect of

sentiment is dominated by the effect of the shortsale constraint for the alternative asset, and

so its price increases. In contrast, the price of the equity index decreases, because beliefs

34

Our computations show that about two-thirds of the trading volume in the model are driven by changes

in the posterior mean, with the remaining one-third being due to changes in the posterior variance.

35

Note also that in the presence of transaction costs, the turnover in the equity index is three to four times

the turnover in the alternative asset; the turnover in the equity index would be even higher if we allowed

for some learning about the dividends of this asset.

26

about its dividends are homogeneous, investors do not wish to short it, and so the positive

price effect of the shortsale constraint is absent.

When we introduce costs for trading the alternative asset, the impact on equilibrium

asset prices is significant. For example, Panel A of Table 4 shows that a transaction cost

of 2.5% leads to a price decline of 2.83% and a transaction cost of 10% leads to a price

decline of 5.17% at the initial date. At later dates, the price discount can be as large as

12.8% for a transaction cost of 10%, in line with the empirical findings of Hege and Nuti

(2011) and Franzoni, Nowak, and Phalippou (2012) for the private-equity market. Contrary

to intuition, the discount is larger in a setting where investors incur modest trading costs,

compared to a setting where there is a blackout and all trade is precluded exogenously. This

is because of the nonmonotone effect of transaction costs on the investors asset-allocation

decisions. Specifically, in the case of full illiquidity, the inexperienced investor buys more

of the asset initially because she cannot trade in the future, which reduces the discount.

In contrast to the price of the alternative asset, we see from Panel B of Figure 4 that

the price of the equity index increases in the cost for trading the alternative asset. The

intuition for this is that if there is a cost for trading the alternative asset, then investors

demand the public-equity index as a substitute, which boosts the price of the equity index.

Qualitatively, the results are the same for the variations from the base case. Noticeably,

the price discount is larger for a higher dividend volatility of the alternative asset and a

lower dividend correlation, because in these cases the diversification loss from tilting the

portfolio is much greater, which increases the price discount. Similarly, when the supply of

the alternative asset is low, the change in demand due to illiquidity has a greater effect on

the price of this asset: The price discount is more than 7% for a transaction cost of 10%,

which would further increase the price discount for asset classes such as hedge funds or small

stocks. Moreover, the longer the horizon of the economy, the stronger is the effect because

the discount applies to more periods. A more precise prior implies smaller tilts in the

portfolio, and hence, a smaller price discount. Finally, the effects of preference parameters

on the price discount of the alternative asset are small (not reported in table).

27

5.6

In a market where both investors have the same level of experience, there is no borrowing

or lending between the two classes of investors. When one of the investors is inexperienced,

she reduces her holding of the alternative asset, and uses the proceeds from the sale of the

alternative asset to buy a little more of the market, with the rest of the funds invested in

the bond because of precautionary savings. This increases the price of the bond and so the

interest rate drops (see Panel C of Table 5). Costs for trading the alternative asset lead to

an increase in bond holdings on average, driven by the desire for precautionary savings (see

Panel A of Figure 5), which leads to a further decrease in the interest rate.

Next, we study the expected returns on the two risky assets. Because the equity index

is traded at each date, its return can be computed in the standard fashion. However, in

the presence of transaction costs, the alternative asset is not traded at each date, and its

return cannot be computed in the standard way. The problem of measuring the returns on

alternative assets such as private equity has been highlighted in several empirical papers;

see, for example, Franzoni, Nowak, and Phalippou (2012), Ang and Sorensen (2012), and

Ang, Papanikolaou, and Westerfield (2014). Franzoni, Nowak, and Phalippou advocate

computing a modified internal rate of return, which measures the average return for an

investor who deposits dividends into an account that earns the interest rate. We, too,

compute the return on the alternative asset using the modified internal rate of return.36

Panel A of Table 5 reports an increase in expected return for the alternative asset as we

increase transaction costs. The observed effects mirror those for the price decrease of the

alternative asset. The magnitudes are again quite large, e.g., for the base case we find an

increase of more than 0.50% compared to the liquid counterpart, which is consistent with

the empirical results in Metrick and Yasuda (2010), who estimate a 1% annual premium

for venture capital. An increase in the transaction cost for the alternative asset increases

demand for the equity index, thereby increasing its price. This price increase of the equity

36

For robustness, we also use a second measure of return, which is computed the same way as the first

measure but assumes reinvestment at a zero interest rate, which removes any equilibrium effects caused by

changes in the endogenous interest rate. Expected returns using these two measure are very similar, and so

we report only the results based on the first measure.

28

index is reflected in a decrease in its expected return; however, as we see from Panel B of

Table 5, the magnitude of this effect is very small.37 The results for the variations in the

base-case parameters mirror those for asset prices, so we do not repeat this discussion.

Next, we study the implications for the equity risk premium. The equity risk premium

increases with opaqueness because the presence of inexperienced investors introduces an

additional risk in the market, for which investors need to be compensated. Transaction

costs reduce both the expected return on the equity index and the interest rate. However,

the effect on the interest rate is always greater, so the net effect on the equity risk premium

is positive although the magnitude is extremely small.

Table 6 presents the impact of opaqueness and illiquidity on return volatilities.38 Intuitively, the trading of inexperienced investors gives rise to an additional source of risk

sentiment riskwhich is reflected in higher volatility for the returns of both risky assets and

also higher correlation between their returns, relative to the perfect-market case. When the

alternative asset can be traded only at a cost, this introduces an additional source of risk,

driving up volatilities further. However, with transaction costs, there is considerably less

information-based trading by the inexperienced investor, which reduces volatility. While

for transaction costs of less than 100% the first effect always dominates, leading to higher

volatility, the second effect typically dominates for the full-blackout cases, thus reducing

the alternative assets return volatility. Finally, transaction costs for trading the alternative

asset have only a negligible effect on the volatility of the equity index.

5.7

Above, our focus has been on the financial policies of the two investors. We now study

the consumption generated by the asset-allocation decisions of the two investors. In particular, we look at the consumption share and the volatility of consumption growth. In

the benchmark perfect-market case, the two investors are identical and have exactly the

37

Note, because of the different procedures for computing the returns of the equity index and the alternative

asset, the two are not directly comparable.

38

For the alternative asset, volatility is measured as the standard deviation across paths of the measure

based on the modified internal rate of return.

29

same consumption share and consumption-growth volatility. However, when one of the

investors is inexperienced, she reduces her holdings of the alternative asset and increases

her holdings of the bond. Consequently, the volatility of her consumption growth decreases

(see Panel A of Table 7). The experienced investor, on the other hand, holds a levered

portfolio with a much larger share of wealth invested in the alternative asset. Therefore,

her consumption-growth volatility is higher than that in the benchmark case.

Of course, this risk-sharing arrangement, where the experienced investor offers insurance to the inexperienced investor, needs to be paid for.39 This payment is reflected in the

transfer of wealth from the inexperienced investor to the experienced one, which over time

leads the experienced investor to consume a greater share of the aggregate dividend. This

effect is illustrated in Panel A of Figure 7. Noticeably, as the inexperienced investor gains

experience over time and there is a reduction in estimation risk (posterior variance), she is

willing to accept more consumption risk (Panels B and C of Figure 7 show the dynamics of

the risk-sharing arrangement). Accordingly, she reduces the premium for the insurance

and, thus, the slope of the change in her consumption share declines (cf. Panel A).40

Focusing now on the effect of transaction costs, Panel A of Figure 7 depicts that the average consumption share of the inexperienced investor monotonically increases with transaction costs on the alternative asset. This is because with transaction costs the inexperienced

investor reduces her portfolio tilt away from the alternative asset. As Table 7 shows, transaction costs have a nonmonotone effect on consumption volatility. This is because of two

opposing forces. First, because of transaction costs, the inexperienced investor reacts less

aggressively to new information, reducing her consumption volatility. Second, transaction

costs introduce a friction into the optimal risk-sharing arrangement, and thus, lead to an increase in consumption-growth volatility. For high transaction costs, especially the case of a

full blackout, the first effect dominates, thereby actually reducing consumption volatility for

the inexperienced investor. Table 7 also shows how variations in the base-case parameters

39

Alternatively, one can attribute this wealth transfer to the inexperienced investors beliefs, which differ

from the objective probability measure.

40

The risk-sharing arrangement between the experienced and inexperienced investors is captured well by

the Brian Tracys quote that: When a man with money meets up with a man with experience, the man

with the experience is going to end up with the money and the man with money is going to end up with the

experience. We are grateful to Francis Longstaff for bringing this quote to our attention.

30

affect the magnitude of these two effects. We also study the dynamics of the consumption

share and consumption-growth volatility for different levels of correlation, volatility, and

precision. Figure 8 shows that a lower correlation and a higher precision imply that the

inexperienced investor is willing to accept a higher initial consumption-growth volatility as

she buys more of the alternative asset. This, in turn, implies a smaller reduction in her

consumption share.

Conclusion

Standard models of asset pricing and asset allocation assume that investors know the distribution of asset returns and that assets can be traded without cost. We study the implications of relaxing these two assumptions simultaneously. Specifically, we consider a

general-equilibrium model of an exchange economy featuring an alternative asset that is

illiquid and opaque, representing private equity, hedge funds, or real assets. The model

features two groups of investors who differ in their experience, with the more experienced

investors having a more precise estimate of the expected growth rate of the alternative

assets dividends.

The interplay between differences in investors experience and differences in the liquidity

of assets has strong implications for asset allocation. When investors differ in experience

only, the inexperienced investors invest less in the alternative asset because of its estimation

risk, tilting their portfolio toward liquid assets, while experienced investors hold a much

greater share of the alternative asset. As inexperienced investors gain experience, they

increase their holdings of the alternative asset. Surprisingly, an increase in the transaction

cost for the alternative asset can lead an investor to hold a larger share of this asset at the

initial date. The intuition for this is that although the investor would ideally like to start

with a low holding of the alternative asset, increasing it over time as she gains experience,

this strategy entails substantial trading costs. Thus, it can be optimal for the inexperienced

investor to reduce her portfolio tilt away from the alternative asset at the initial date, and

thus, hold a larger share of this asset in order to economize on future trading costs. Another

31

striking result is that, because of the portfolio inertia induced by transaction costs, the

inexperienced investor might actually hold a majority of the alternative asset even though

she is currently pessimistic about its growth rate. We also find interesting results for the

inexperienced investors holdings of the liquid assets. If investors find it optimal not to

trade the alternative asset because of its transaction costs, then investors trade the liquid

equity index instead, leading to strong spillover effects.

Similarly, experience and illiquidity have sizable effects on asset prices. In the presence of

transaction costs for the alternative asset, its price drops substantially: for a 5% transaction

cost, the average price discount is about 4%, with a maximum of more than 12%. Contrary

to intuition, the illiquidity discount is stronger for moderate transaction costs than for full

illiquidity. In contrast, prices for the liquid assets increase with transaction costs for the

alternative asset, though this increase is smaller in magnitude.

These results are consistent with various empirical observations about alternative assets.

For example, several papers (Lerner, Schoar, and Wang (2008) and Goetzman and Oster

(2012), among others) document that large endowment funds have increased their exposure

to alternative assets over the last years after experiencing positive returns and today strongly

overweight alternative assetsrelative to the market portfolio and relative to their lessexperienced peers, confirming our static and dynamic asset-allocation results. Similarly,

our asset pricing implications with an illiquidity discount of 5%10% are comparable to

results reported by Hege and Nuti (2011) and Franzoni, Nowak, and Phalippou (2012) for

private equity.

We find several novel results for asset allocation and asset pricing, originating from

incorporating both illiquidity and experiential learning into our model, characteristics that

are absent from traditional models. Specifically, our work demonstrates how inexperienced

investors should allocate wealth to new assets that are illiquid and opaque, how they should

revise their investments as they gain experience, and what the pricing implications are for

both new and existing assets.

32

First-Order Conditions

The Lagrangian associated with equations (4), (5), and the shortsale constraint is:

L(k, t) =

sup

inf

"

+ BC (k, t) (0, k, t 1) +

h

1 i

1

1

k

1 1

(1 ) c(k, t)

+ Et V (k, t + 1)

2

X

(n, k, t 1) S(n, k, t) + d(n, t)

n=1

c(k, t)

2

X

#

(n, k, t)S(n, k, t) (2, k, t), (2, k, t 1), S(2, k, t)

n=0

(A1)

where BC (k, t) and BC (k, t)SSC (k, t)S(2, k, t) denote the Lagrange multipliers associated

with the budget constraint and the shortsale constraint, respectively.

The first-order condition of (A1) with respect to consumption c(k, t) is given by:

L(k, t)

1

=

c(k, t)

1

(

1

1

1

(1 ) c(k, t)

+ Et V (k, t + 1)

1

1

1

1

1 1

1

1 1 1

(1 ) 1

c(k, t) BC (k, t)

= (1 ) c(k, t)

= (1 ) c(k, t)

(

) 1 1 1

1 1

1

1

BC (k, t)

(1 ) c(k, t)1 + Et V (k, t + 1)1 1

1

V (k, t) BC (k, t) 0.

(A2)

Second, the first-order condition of (A1) with respect to the Lagrange multiplier BC (k, t):

"

2

2

X

X

L(k, t)

= (0, k, t 1) +

(n, k, t 1) S(n, k, t) + d(n, t)

(n, k, t)S(n, k, t)

BC (k, t)

n=1

n=0

33

i

Third, the derivative of (A1) with respect to the holdings of the first two assets (n {0, 1}):

L(k, t)

(n, k, t)

1

1

1

1

(1 ) c(k, t)

k

+ Et V (k, t + 1)

1

1

1

1 1

1

1 1

V (k, t + 1)

Etk V (k, t + 1)1 1 (1 ) Etk V (k, t + 1)

1

(n, k, t)

BC (k, t)S(n, k, t)

1

= V (k, t)

Etk

V (k, t + 1)

1 1

1

Etk

V (k, t + 1)

V (k, t + 1)

(n, k, t)

BC (k, t)S(n, k, t) 0.

Finally, taking the derivative of (A1) with respect to the holdings of the illiquid asset:

L(k, t)

(2, k, t)

1

1

(

(1 ) c(k, t)

1

1

Etk

1

1 1

V (k, t + 1)

1

1 1

1 1

V (k, t + 1)

V (k, t + 1)

(1

V (k, t + 1)

1

(2, k, t)

!

(2, k, t), (2, k, t 1), S(2, k, t)

BC (k, t) S(2, k, t) +

(2, k, t)

Etk

) Etk

1

1

1 1 1 k

V (k, t + 1)

V (k, t + 1)

Et V (k, t + 1)

= V (k, t)

(2, k, t)

!

(2, k, t), (2, k, t 1), S(2, k, t)

BC (k, t) S(2, k, t) +

(2, k, t)

1

Etk

To summarize, the Karush-Kuhn-Tucker first-order conditions are:

BC (k, t) = (1 ) c(k, t)

0 = (0, k, t 1) +

V (k, t) ,

2

X

(A3)

(n, k, t 1) S(n, k, t) + d(n, t)

n=1

c(k, t)

2

X

(n, k, t)S(n, k, t) (2, k, t), (2, k, t 1), S(2, k, t) , (A4)

n=0

34

V (k, t) k

S(n, k, t) =

Et V (k, t + 1)1 1

BC (k, t)

V (k, t + 1)

Etk V (k, t + 1)

, n {0, 1} ,

(n, k, t)

(A5)

V (k, t)

S(2, k, t) =

1

(k, t + 1)

V (k, t + 1)

V (k, t + 1)

(2, k, t)

(2,

k,

t),

(2,

k,

t

1),

S(2,

k,

t)

1

,

1 SSC (k, t)

(2, k, t)

Etk

(2, k, t) 0;

SSC (k, t) 0;

Etk

(A6)

(A7)

Equation (A3) equates the marginal utility of consumption to BC (k, t), the shadow price

for relaxing the budget constraint. Equation (A4) is the budget constraint that the consumption and portfolio policies must satisfy. Equations (A5) and (A6) equate the benefit from holding an asset to the benefit from selling itfor the alternative asset, net of

transaction costs and factoring in the shortsale constraint. Finally, equations (A7) are the

complementary slackness and inequality conditions associated with the shortsale constraint.

From the envelope theorem we know that we can replace the derivative of the value

function with the partial derivative of the Lagrangian, assuming that the decision variables

are at their optimum and hence the first-order conditions with respect to decision variables

are satisfied. Differentiating the original Lagrangian (A1) with respect to past holdings we

get

L(k, t)

V (k, t)

=

= BC (k, t) 1,

(0, k, t 1)

(0, k, t 1)

(A8)

V (k, t)

L(k, t)

=

= BC (k, t) S(1, k, t) + d(1, t) ,

(1, k, t 1)

(1, k, t 1)

(A9)

V (k, t)

L(k, t)

=

= BC (k, t)

(2, k, t 1)

(2, k, t 1)

S(2, k, t) + d(2, t)

!

(2, k, t), (2, k, t 1), S(2, k, t)

,

(2, k, t 1)

35

(A10)

such that one can rewrite the kernel conditions (A5) and (A6) by replacing the marginal

value function with respect to asset holdings from (A8)(A10) (shifted one period forward), and recognizing from the first-order conditions with respect to consumption (A2)

1

M (k, t + 1)

,

S(0, k, t) = Etk

M (k, t)

k M (k, t + 1)

S(1, k, t) = Et

S(1, k, t + 1) + d(1, t + 1) ,

M (k, t)

(A11)

(A12)

|(2, k, t) (2, k, t 1)| 1

(A13)

1 SSC (k, t) +

(2, k, t)

"

!#

M

(k,

t

+

1)

|(2,

k,

t

+

1)

(2,

k,

t)|

Etk

S(2, k, t + 1) 1

+d(2, t + 1) ,

M (k, t)

(2, k, t)

S(2, k, t) =

where we used the transaction cost function in equation (1) to simplify its derivative, and

where

M (k,t+1)

M (k,t)

1

M (k, t + 1)

=

M (k, t)

1)1

V (k, t +

k

Et [V (k, t + 1)1 ]

1

c(k, t + 1)

c(k, t)

1

As explained in the main text, in a general-equilibrium setting when the investor attempts

to solve for the optimal consumption and portfolio policies at date t, asset prices need to

adjust in order for markets to clear. However, if one were solving the system of equations

backward, these prices could not adjust because they depend on future consumption, which

has already been determined in the previous step. Dumas and Lyasoff (2012) address this

problem by proposing a time-shift whereby at date t one solves for the optimal portfolio

for date t but the optimal consumption for date t+1, instead of the optimal consumption for

date t. Using this insight allows one to write the system of equations so that it is recursive.

Assuming that there are K future states (nodes), the final system of equations to solve

is given by both investors (k {1, 2}) flow budget equations for the next period for all K

36

c(k, t + 1) +

2

X

(n, k, t + 1)S(n, k, t + 1) + (2, k, t + 1), (2, k, t), S(2, k, t + 1)

n=0

= (0, k, t) +

2

X

(n, k, t) S(n, k, t + 1) + d(n, t + 1) ,

(B1)

n=1

together with the complementary slackness conditions and inequality conditions given in

(A7), the kernel conditions given in (A11) to (A13), and the market-clearing conditions for

all three assets:

ss (n) =

2

X

(B2)

k=1

Not counting the inequality conditions, these are 2K + 2 + 3 + 3 equations (2K flow

budget equations, 2 complementarity-slackness conditions for the two investors, 3 kernel

conditions for the prices of the three assets, and 3 financial-market-clearing conditions).

The unknowns are next periods consumption c(k, t + 1) for both investors for all K states,

both investors current periods optimal asset holdings (n, k, t) for the three assets, together

with each investors Lagrange multiplier SSC (k, t) associated with the shortsale constraint;

in total there are 2K + 6 + 2 unknownsthe same number as the number of equations.

This system of equations is solved recursively based on interpolating functions for the

future prices, S(n, k, t+1), next periods utility functions, V (k, t+1), as well as next periods

asset holdings, (n, k, t + 1), using the terminal conditions S(n, k, T ) = (n, k, T ) = 0 for

all states, all assets n, and all investors k.

After solving the system for all dates t {1, ..., T 1}, one has solved all equations from

the global systemexcept the budget equations for the initial date, which have not been

used so far because of the time shift. So, at the initial date, we need to solve the following

system of equations for the two unknowns c(k, 1), k {1, 2}:

c(k, 1) +

2

X

n=0

2

X

(n, k, 0) S(n, k, 1) + d(n, 1) ,

n=1

41

Note, for ease of notation, we have suppressed the dependence of time t + 1 variables on the state

(t + 1, s) in the equations that follow.

37

based on interpolating functions for the initial-date prices, S(n, k, 1), and holdings, (n, k, 1).

The endowed holdings (n, k, 0) for the three assets are exogenous to the system and reflect

the initial periods incoming (endowed) wealth of the two investors.

38

Variable

Description

Base Case

Variation

15

10; 20

Horizon

Rate of time-preference

0.95

Relative risk-aversion

1.75

1.5; 2.0

1.5

(0, k, 0)

(2, k, 0)

0.5

1 , 2

1/1

2/2

D

1

D

2

1D

2D

1.2%

1.2%

20%

20%

12.5%; 27.5%

0.5

0.25; 0.75

0.2

0.1; 0.3

d(1,1)

d(1,1)+d(2,1)

The table reports the model parameters used for our base case as well as the variations in the model

parameters that we considered.

39

Parameter Values

T

Volatility

Altern.

Corr.

Size

Altern.

Initial Holdings

Prior

Perfect

Market

Opaque

Constr.

Relative to Opaque Setup

2.5% TC

5% TC

10% TC

100% TC

15

10

20

20%

12.5%

27.5%

0.5

0.25

0.75

0.2

0.1

0.3

1/1

2/2

0.500

0.500

0.500

0.500

0.500

0.500

0.500

0.500

0.500

0.500

0.123

0.240

0.022

0.006

0.152

0.269

0.000

0.000

0.206

0.243

0.082

0.064

0.022

0.006

0.027

0.027

0.000

0.000

0.082

0.035

0.115

0.035

0.022

0.006

0.024

0.040

0.000

0.000

0.056

0.004

+

+

+

+

0.021

0.009

0.019

0.006

0.075

0.045

0.000

0.000

0.022

0.030

+

+

+

+

+

+

+

+

+

+

0.150

0.083

0.215

0.118

0.191

0.133

0.082

0.281

0.066

0.069

0.2

0.1

0.3

1/1

2/2

0.500

0.500

0.500

0.500

0.500

0.500

0.500

0.500

0.500

0.500

0.508

0.504

0.514

0.510

0.509

0.512

0.484

0.494

0.509

0.504

+

+

0.007

0.003

0.007

0.011

0.003

0.003

0.010

0.007

0.013

0.001

0.007

0.009

0.015

0.017

0.011

0.006

0.014

0.011

0.006

0.013

0.032

0.025

0.025

0.022

0.047

0.027

0.017

0.018

0.029

0.022

0.066

0.040

0.089

0.041

0.088

0.046

0.037

0.053

0.068

0.032

1/1

2/2

0.000

0.000

0.000

0.000

0.000

0.000

0.000

0.000

0.000

0.000

0.736

0.370

1.106

1.089

0.575

0.442

1.075

0.621

0.818

0.505

+

+

+

+

+

+

+

+

+

+

0.092

0.069

0.083

0.063

0.052

0.062

0.054

0.038

0.134

0.069

+

+

+

+

+

+

+

+

+

+

0.143

0.087

0.128

0.093

0.095

0.108

0.077

0.063

0.168

0.072

+

+

+

+

+

+

+

+

+

+

0.145

0.102

0.183

0.122

0.181

0.066

0.095

0.100

0.211

0.088

+

+

+

+

+

+

+

+

+

0.142

0.082

0.192

0.017

0.283

0.019

0.082

0.087

0.215

0.079

15

10

20

20%

12.5%

27.5%

0.5

0.25

0.75

15

10

20

20%

12.5%

27.5%

0.5

0.25

0.75

0.2

0.1

0.3

The table presents the change in optimal initial holdings of the inexperienced investor at the initial

date t = 1 for different levels of transaction costs. Panel A reports the optimal holdings for the

alternative asset, Panel B the optimal holdings for the equity index, and Panel C the optimal

holdings for the risk-free bond. The first line of each panel refers to the base case; the remaining

lines capture variations in parameters. T denotes the economys horizon; Volatility Altern. denotes

the alternative assets dividend volatility; Corr. denotes the correlation between the dividends of the

equity index and the alternative asset; Size Altern. denotes the initial share of the alternative assets

dividends relative to total dividends; Prior denotes the prior of the inexperienced investor, given by

the two hyperparameters of the Beta distribution. The column titled Perfect Market reports the

optimal initial holdings when both risky assets can be traded without costs and both investors have

perfect knowledge. The column titled Opaque Constr. reports the optimal initial holdings when both

risky assets can be traded without costs, but the expected growth rate of dividends is opaque for

the inexperienced investor, and both investors face shortsale constraints.

40

Parameter Values

T

Volatility

Altern.

Corr.

Size

Altern.

Avg. Turnover

Prior

Perfect

Market

Opaque

Constr.

Relative to Opaque Setup

2.5% TC

5% TC

10% TC

100% TC

15

10

20

20%

12.5%

27.5%

0.5

0.25

0.75

0.2

0.1

0.3

1/1

2/2

0.000

0.000

0.000

0.000

0.000

0.000

0.000

0.000

0.000

0.000

0.627

0.498

0.697

0.738

0.535

0.600

0.632

0.375

0.809

0.535

42.17

38.92

47.76

51.79

41.00

45.10

52.84

41.73

45.53

56.23

49.80

56.72

56.10

61.20

51.42

57.20

60.70

51.14

58.65

68.22

69.11

72.99

64.51

70.93

68.54

74.56

70.62

60.86

71.11

80.30

89.47

88.52

90.15

83.58

92.47

95.03

81.80

91.03

88.25

89.80

0.2

0.1

0.3

1/1

2/2

0.000

0.000

0.000

0.000

0.000

0.000

0.000

0.000

0.000

0.000

0.463

0.350

0.513

0.507

0.407

0.256

0.662

0.535

0.376

0.278

+

+

+

+

+

+

+

+

+

+

28.01

33.19

28.06

24.47

35.43

14.80

40.33

19.21

52.41

51.70

+

+

+

+

+

+

+

+

+

+

35.95

46.12

33.76

27.23

47.17

25.94

45.76

25.30

61.60

61.93

+

+

+

+

+

+

+

+

+

+

44.67

59.43

36.67

27.29

69.13

35.88

50.84

28.47

74.08

72.29

+

+

+

+

+

+

+

+

+

+

58.48

70.72

55.03

35.26

89.82

44.40

56.54

41.51

90.16

79.25

1/1

2/2

0.000

0.000

0.000

0.000

0.000

0.000

0.000

0.000

0.000

0.000

1.063

0.890

1.146

1.274

0.898

0.805

1.319

0.950

1.151

0.803

8.68

8.42

9.31

18.04

1.16

20.59

3.04

5.21

11.58

14.58

15

10

20

20%

12.5%

27.5%

0.5

0.25

0.75

15

10

20

20%

12.5%

27.5%

0.5

0.25

0.75

0.2

0.1

0.3

10.40

12.49

10.74

22.38

0.59

26.31

4.08

6.08

14.68

18.42

14.57

16.83

13.66

28.37

0.15

34.71

6.53

7.99

18.80

23.53

21.05

21.40

21.46

33.34

4.67

47.21

9.15

12.80

25.91

27.61

The table presents the percentage change in average turnover (average number of shares traded

between t = 1 and T ) for different levels of transaction costs. Panel A reports the average turnover

for the alternative asset, Panel B the average turnover for the equity index, and Panel C the average

turnover for the risk-free bond. The first line of each panel refers to the base case; the remaining

lines capture variations in parameters. T denotes the economys horizon; Volatility Altern. denotes

the alternative assets dividend volatility; Corr. denotes the correlation between the dividends of the

equity index and the alternative asset; Size Altern. denotes the initial share of the alternative assets

dividends relative to total dividends; Prior denotes the prior of the inexperienced investor, given

by the two hyperparameters of the Beta distribution. The column titled Perfect Market reports

the average turnover when both risky assets can be traded without costs and both investors have

perfect knowledge. The column titled Opaque Constr. reports the average turnover when both risky

assets can be traded without costs, but the expected growth rate of dividends is opaque for the

inexperienced investor, and both investors face shortsale constraints.

41

Parameter Values

T

Volatility

Altern.

Corr.

Size

Altern.

Asset Price

Prior

Perfect

Market

Opaque

Constr.

Relative to Opaque Setup

2.5% TC

5% TC

10% TC

100% TC

15

10

20

20%

12.5%

27.5%

0.5

0.25

0.75

0.2

0.1

0.3

1/1

2/2

2.059

1.466

2.528

2.259

1.837

2.155

1.958

1.085

2.974

2.059

2.125

1.509

2.599

2.351

1.878

2.269

1.984

1.162

3.016

2.083

2.83

2.53

2.78

2.43

2.85

3.26

1.85

3.21

2.45

1.99

4.37

3.46

4.31

3.50

4.16

4.74

2.60

5.20

3.35

2.65

5.17

4.12

6.11

4.33

5.33

6.19

3.25

7.54

4.02

2.91

4.41

3.54

4.73

3.86

4.41

5.51

2.91

7.28

3.10

2.46

0.2

0.1

0.3

1/1

2/2

7.310

5.373

8.743

7.187

7.453

7.274

7.353

8.237

6.427

7.310

7.286

5.360

8.708

7.131

7.465

7.262

7.316

8.186

6.433

7.298

+

+

+

+

+

+

+

+

+

+

0.07

0.04

0.06

0.10

0.04

0.04

0.09

0.03

0.09

0.02

+

+

+

+

+

+

+

+

+

+

0.17

0.09

0.17

0.24

0.05

0.06

0.16

0.10

0.11

0.02

+

+

+

+

+

+

+

+

+

+

0.17

0.16

0.34

0.40

0.06

0.11

0.30

0.22

0.11

0.06

+

+

+

+

+

+

+

+

+

+

0.41

0.38

0.36

0.61

0.19

0.45

0.40

0.49

0.26

0.16

15

10

20

20%

12.5%

27.5%

0.5

0.25

0.75

The table presents the percentage change in initial stock prices, as we vary transaction costs. Panel A

reports the initial stock price for the alternative asset and Panel B the initial stock price for the

equity index. The first line of each panel refers to the base case; the remaining lines capture

variations in parameters. T denotes the economys horizon; Volatility Altern. denotes the alternative

assets dividend volatility; Corr. denotes the correlation between the dividends of the equity index

and the alternative asset; Size Altern. denotes the initial share of the alternative assets dividends

relative to total dividends; Prior denotes the prior of the inexperienced investor, given by the two

hyperparameters of the Beta distribution. The column titled Perfect Market reports the initial stock

price when both risky assets can be traded without costs and both investors have perfect knowledge.

The column titled Opaque Constr. reports the initial stock price when both risky assets can be

traded without costs, but the expected growth rate of dividends is opaque for the inexperienced

investor, and both investors face shortsale constraints.

42

Parameter Values

T

Volatility

Altern.

Corr.

Size

Altern.

Expected Return

Prior

Perfect

Market

Opaque

Constr.

Relative to Opaque Setup

2.5% TC

5% TC

10% TC

100% TC

15

10

20

20%

12.5%

27.5%

0.5

0.25

0.75

0.2

0.1

0.3

1/1

2/2

4.295

4.051

4.618

3.734

5.055

4.139

4.507

3.474

4.869

4.263

4.265

3.901

4.727

3.461

5.367

3.914

4.649

3.034

5.038

4.385

+

+

+

+

+

+

+

+

+

+

0.301

0.364

0.215

0.236

0.333

0.346

0.188

0.326

0.262

0.243

+

+

+

+

+

+

+

+

+

+

0.459

0.492

0.344

0.324

0.482

0.504

0.251

0.530

0.344

0.309

+

+

+

+

+

+

+

+

+

+

0.528

0.569

0.481

0.363

0.611

0.665

0.260

0.768

0.395

0.313

+

+

+

+

+

+

+

+

+

+

0.344

0.411

0.258

0.274

0.353

0.535

0.150

0.676

0.178

0.170

0.2

0.1

0.3

1/1

2/2

7.512

7.553

7.498

7.829

7.165

7.567

7.431

7.506

7.408

7.504

7.639

7.672

7.631

8.015

7.256

7.655

7.652

7.667

7.481

7.584

0.016

0.017

0.026

0.019

0.012

0.018

0.012

0.009

0.018

0.000

0.031

0.029

0.038

0.039

0.018

0.028

0.021

0.019

0.027

0.002

0.035

0.043

0.059

0.064

0.023

0.037

0.040

0.037

0.027

0.006

0.071

0.097

0.052

0.092

0.049

0.075

0.052

0.076

0.053

0.035

1/1

2/2

1.100

1.099

1.100

1.750

0.432

1.562

0.645

0.745

1.328

1.099

0.826

0.938

0.750

1.545

0.096

1.539

0.061

0.532

0.972

0.892

0.027

0.027

0.027

0.032

0.018

0.020

0.033

0.010

0.033

0.011

0.057

0.044

0.053

0.070

0.029

0.031

0.056

0.030

0.052

0.018

0.066

0.062

0.094

0.110

0.036

0.042

0.092

0.064

0.061

0.025

0.064

0.090

0.046

0.112

0.009

0.041

0.089

0.069

0.045

0.040

15

10

20

20%

12.5%

27.5%

0.5

0.25

0.75

15

10

20

20%

12.5%

27.5%

0.5

0.25

0.75

0.2

0.1

0.3

The table presents the change in percentage expected returns for different levels of transaction

costs. Panel A reports the modified internal rate of return for the alternative asset, Panel B the

expected return for the equity index, and Panel C the risk-free rate for the bond. The first line

of each panel refers to the base case; the remaining lines capture variations in parameters. T

denotes the economys horizon; Volatility Altern. denotes the alternative assets dividend volatility;

Corr. denotes the correlation between the dividends of the equity index and the alternative asset;

Size Altern. denotes the initial share of the alternative assets dividends relative to total dividends;

Prior denotes the prior of the inexperienced investor, given by the two hyperparameters of the Beta

distribution. The column titled Perfect Market reports the percentage expected returns when both

risky assets can be traded without costs and both investors have perfect knowledge. The column

titled Opaque Constr. reports the percentage expected returns when both risky assets can be traded

without costs, but the expected growth rate of dividends is opaque for the inexperienced investor,

and both investors face shortsale constraints.

43

Parameter Values

T

Volatility

Altern.

Corr.

Size

Altern.

Return Volatility

Prior

Relative to Opaque Setup

Perfect

Market

Opaque

Constr.

2.5% TC

5% TC

10% TC

100% TC

15

10

20

20%

12.5%

27.5%

0.5

0.25

0.75

0.2

0.1

0.3

1/1

2/2

5.566

5.889

5.721

3.158

8.487

5.515

5.657

5.196

5.741

5.576

6.647

6.702

7.164

3.769

10.548

6.308

7.063

5.805

7.097

6.535

+

+

+

+

+

+

+

+

+

+

0.156

0.154

0.145

0.077

0.241

0.185

0.092

0.184

0.112

0.073

+

+

+

+

+

+

+

+

+

+

0.231

0.200

0.214

0.096

0.335

0.265

0.113

0.297

0.134

0.095

+

+

+

+

+

+

+

+

+

+

0.233

0.181

0.266

0.042

0.391

0.333

0.033

0.414

0.099

0.037

0.191

0.053

0.410

0.114

0.462

0.017

0.292

0.158

0.393

0.181

0.2

0.1

0.3

1/1

2/2

21.194

21.127

21.188

21.358

21.075

21.154

21.188

21.242

21.088

21.185

21.590

21.445

21.646

21.921

21.395

21.155

22.317

21.720

21.412

21.454

0.015

0.014

0.015

0.007

0.016

0.021

0.003

0.008

0.016

0.002

0.015

0.007

0.016

0.001

0.020

0.027

0.021

0.008

0.012

0.008

+

+

+

+

+

+

0.009

0.022

0.004

0.045

0.010

0.020

0.085

0.006

0.018

0.031

+

+

+

+

+

+

+

+

+

+

0.077

0.053

0.088

0.091

0.035

0.009

0.165

0.051

0.095

0.044

15

10

20

20%

12.5%

27.5%

0.5

0.25

0.75

The table presents the change in percentage volatility for different levels of transaction costs. Panel A

reports the dispersion across paths in the modified internal rate of return for the alternative asset

and Panel B the volatility for the equity index. The first line of each panel refers to the base case;

the remaining lines capture variations in parameters. T denotes the economys horizon; Volatility

Altern. denotes the alternative assets dividend volatility; Corr. denotes the correlation between the

dividends of the equity index and the alternative asset; Size Altern. denotes the initial share of the

alternative assets dividends relative to total dividends; Prior denotes the prior of the inexperienced

investor, given by the two hyperparameters of the Beta distribution. The column titled Perfect

Market reports the percentage volatility when both risky assets can be traded without costs and

both investors have perfect knowledge. The column titled Opaque Constr. reports the percentage

volatility when both risky assets can be traded without costs, but the expected growth rate of

dividends is opaque for the inexperienced investor, and both investors face shortsale constraints.

44

Parameter Values

T

Volatility

Altern.

Corr.

Size

Altern.

Cons. Volatility

Prior

Relative to Opaque Setup

Perfect

Market

Opaque

Constr.

2.5% TC

5% TC

10% TC

1/1

2/2

5.566

5.889

5.721

3.158

8.487

5.515

5.657

5.196

5.741

5.576

6.647

6.702

7.164

3.769

10.548

6.308

7.063

5.805

7.097

6.535

+

+

+

+

+

+

+

+

+

+

0.156

0.154

0.145

0.077

0.241

0.185

0.092

0.184

0.112

0.073

+

+

+

+

+

+

+

+

+

+

0.231

0.200

0.214

0.096

0.335

0.265

0.113

0.297

0.134

0.095

+

+

+

+

+

+

+

+

+

+

0.233

0.181

0.266

0.042

0.391

0.333

0.033

0.414

0.099

0.037

0.191

0.053

0.410

0.114

0.462

0.017

0.292

0.158

0.393

0.181

1/1

2/2

21.194

21.127

21.188

21.358

21.075

21.154

21.188

21.242

21.088

21.185

21.590

21.445

21.646

21.921

21.395

21.155

22.317

21.720

21.412

21.454

0.015

0.014

0.015

0.007

0.016

0.021

0.003

0.008

0.016

0.002

0.015

0.007

0.016

0.001

0.020

0.027

0.021

0.008

0.012

0.008

+

+

+

+

+

+

0.009

0.022

0.004

0.045

0.010

0.020

0.085

0.006

0.018

0.031

+

+

+

+

+

+

+

+

+

+

0.077

0.053

0.088

0.091

0.035

0.009

0.165

0.051

0.095

0.044

100% TC

15

10

20

20%

12.5%

27.5%

0.5

0.25

0.75

0.2

0.1

0.3

15

10

20

20%

12.5%

27.5%

0.5

0.25

0.75

0.2

0.1

0.3

The table presents the change in percentage consumption volatility for the two investors for different

levels of transaction costs. Panel A reports the consumption volatility for the inexperienced investor

and Panel B for the experienced investor. The first line of each panel refers to the base case;

the remaining lines capture variations in parameters. T denotes the economys horizon; Volatility

Altern. denotes the alternative assets dividend volatility; Corr. denotes the correlation between the

dividends of the equity index and the alternative asset; Size Altern. denotes the initial share of the

alternative assets dividends relative to total dividends; Prior denotes the prior of the inexperienced

investor, given by the two hyperparameters of the Beta distribution. The column titled Perfect

Market reports the percentage consumption volatility when both risky assets can be traded without

costs and both investors have perfect knowledge. The column titled Opaque Constr. reports the

percentage consumption volatility when both risky assets can be traded without costs, but the

expected growth rate of dividends is opaque for the inexperienced investor, and both investors face

shortsale constraints.

45

Dividend

0.6

0.275

0.45

0.175

2

10

12

Inexperienced

Investors Posterior

0.75

0.375

0.3

14

Time

Dividend (left yaxis)

Dividend

0.6

0.275

0.45

0.175

2

10

12

0.3

14

12

14

Inexperienced

Investors Posterior

0.75

0.375

Time

Panel C: Simulations 1&2 for the Equity Index

Dividend

1.35

1.15

0.95

0.75

2

10

Time

The figure shows the evolution of the dividends for the alternative asset and the equity index over

time for two simulated paths of the economy. Panels A and B plot the dividend dynamics of

the alternative asset (left-hand-side vertical scale), together with the resulting posterior mean and

the scaled posterior standard deviation for the probability of an up-move in the alternative assets

dividend tree (both on right-hand-side vertical scale), as perceived by the inexperienced investor.

The shaded gray areas indicate periods of negative dividend realizations. Panel C depicts the

evolution of the dividends of the equity index, coinciding for both paths. Simulations are based on

the parameter values for the base case.

46

0.25

Frequency

0.2

0.15

0.1

0.05

0.2

0.4

0.6

0.8

Posterior Mean

The figure shows the frequency diagram for the inexperienced investors posterior mean of the

probability of an up-move at date T = 15, as given in equation (2). The computations are based on

a Beta(1,1) prior and 50,000 simulated paths of the underlying dividend tree.

47

0.5

0.4

0.3

0.2

0.1

0

10

12

14

Time

Panel B: Alternative Asset Holdings of the Inexperienced Investor in Simulation 1

1

0.8

0.6

0.4

0.2

2

10

12

14

Time

Panel C: Alternative Asset Holdings of the Inexperienced Investor in Simulation 2

1

0.8

0.6

0.4

0.2

2

10

12

14

Time

2.5%

10%

100%

The figure plots the holdings of the inexperienced investor in the alternative asset over time. Panel A

shows the average holdings (across paths), whereas Panels B and C depict the holdings for the two

simulated paths of the economy, which are described in Figure 1. The shaded gray areas indicate

periods of negative dividend realizations for the alternative asset. Results are based on the parameter

values for the base case and under the assumption that the alternative asset is opaque. The dynamics

are shown for transaction cost levels of 2.5%, 10%, and 100%.

48

0.54

0.52

0.5

0.48

0.46

0.44

10

12

14

Time

Panel B: Equity Index Holdings of the Inexperienced Investor in Simulation 1

0.8

0.7

0.6

0.5

0.4

2

10

12

14

Time

Panel C: Equity Index Holdings of the Inexperienced Investor in Simulation 2

0.8

0.7

0.6

0.5

0.4

2

10

12

14

Time

2.5%

10%

100%

The figure plots the holdings of the inexperienced investor in the equity index over time. Panel A

shows the average holdings (across paths), whereas Panels B and C depict the holdings for the two

simulated paths of the economy, which are described in Figure 1. The shaded gray areas indicate

periods of negative dividend realizations for the alternative asset. Results are based on the parameter

values for the base case and under the assumption that the alternative asset is opaque. The dynamics

are shown for transaction cost levels of 2.5%, 10%, and 100%.

49

1

0.5

0.5

10

12

14

Time

Panel B: Bond Holdings of the Inexperienced Investor in Simulation 1

0.5

0

0.5

1

1.5

2

2.5

2

10

12

14

Time

Panel C: Bond Holdings of the Inexperienced Investor in Simulation 2

0.5

0

0.5

1

1.5

2

2.5

2

10

12

14

Time

2.5%

10%

100%

The figure plots the holdings of the inexperienced investor in the bond over time. Panel A shows the

average holdings (across paths), whereas Panels B and C depict the holdings for the two simulated

paths of the economy, which are described in Figure 1. The shaded gray areas indicate periods of

negative dividend realizations for the alternative asset. Results are based on the parameter values

for the base case and under the assumption that the alternative asset is opaque. The dynamics are

shown for transaction cost levels of 2.5%, 10%, and 100%.

50

Panel A: Alternative Asset Holdings of the Inexperienced Investor for Simulation 1 with 2.5% TC

0.8

0.6

0.4

0.2

10

12

14

Time

Panel B: Equity Index Holdings of the Inexperienced Investor for Simulation 1 with 2.5% TC

0.65

0.6

0.55

0.5

0.45

2

10

12

14

Time

Panel C: Bond Holdings of the Inexperienced Investor for Simulation 1 with 2.5% TC

0.5

0.5

1

2

10

12

14

Time

Base Case

Lower Correlation

Higher Volatility

Higher Precision

The figure shows the dynamics of the asset allocation of the inexperienced investor for a simulated

path of the economy, which is described in Panel A of Figure 1, for various alternative specifications

of the parameter values. Panel A shows the holdings in the alternative asset, Panel B the holdings

in the equity index, and Pane C the holdings in the risk-free bondin each case plotted over time.

The shaded gray areas indicate periods of negative dividend realizations for the alternative asset.

Results are shown for the base case, the case in which the alternative assets dividend growth rate

has a volatility of 27.5% (Higher Volatility), the case in which the correlation between the risky

assets dividends is 0.25 (Lower Correlation), and the case in which the inexperienced investor

starts out with a more precise prior (Higher Precision). All results are for a transaction cost of

2.5%.

51

0.54

0.52

0.5

0.48

0.46

0.44

10

12

14

Time

Panel B: Average Consumption Growth Volatility of the Inexperienced Investor

0.24

0.22

0.2

0.18

0.16

0.14

2

10

12

14

Time

Panel C: Average Consumption Growth Volatility of the Experienced Investor

0.24

0.22

0.2

0.18

0.16

0.14

2

10

12

14

Time

2.5%

10%

100%

The figure shows the dynamics of the two investors consumption growth over time. Panel A shows

the evolution of the inexperienced investors average consumption share. Panels B and C show the

evolution of the volatility of consumption growth faced by the inexperienced and the experienced

investors, respectively. Results are based on the parameters values for the base case and under the

assumption that the alternative asset is opaque. The dynamics are shown for transaction cost levels

of 2.5%, 10%, and 100%.

52

Values

0.52

0.5

0.48

0.46

0.44

2

10

12

14

Time

Panel B: Average Consumption Growth Volatility of the Inexperienced Investor with 2.5% TC

0.2

0.18

0.16

0.14

2

10

12

14

Time

Panel C: Average Consumption Growth Volatility of the Experienced Investor with 2.5% TC

0.28

0.26

0.24

0.22

0.2

0.18

2

10

12

14

Time

Base Case

Lower Correlation

Higher Volatility

Higher Precision

The figure shows the consumption growth dynamics for alternative specifications of the parameter

values. Panel A shows the evolution of the inexperienced investors average consumption share.

Panels B and C show the evolution of the volatility of consumption growth faced by the inexperienced

and the experienced investor. Results are shown for the base case, the case in which the alternative

assets dividend growth rate has a volatility of 27.5% (Higher Volatility), the case in which the

correlation between the risky assets dividends is 0.25 (Lower Correlation), as well as the case

in which the inexperienced investor starts out with a more precise prior (Higher Precision). All

results are for a transaction cost of 2.5%.

53

References

Acharya, V. V., and L. H. Pedersen, 2005, Asset Pricing with Liquidity Risk, Journal of

Financial Economics, 77, 375410.

Amihud, Y., H. Mendelson, and L. H. Pedersen, 2005, Liquidity and Asset Prices, Foundations and Trends in Finance, 1(4), 269364.

Ang, A., 2013, Illiquid Asset Investing, Working Paper, Columbia University.

Ang, A., A. Ayala, and W. N. Goetzman, 2014, Investment Beliefs of Endowments,

Working Paper, Columbia University and Yale University.

Ang, A., D. Papanikolaou, and M. M. Westerfield, 2014, Portfolio Choice with Illiquid

Assets, Management Science, 60(11), 27372761.

Ang, A., and M. Sorensen, 2012, Risks, Returns, and Optimal Holdings of Private Equity:

A Survey of Existing Approaches, Quarterly Journal of Finance, 2(3), 1250011.

, 2013, Investing in Private Equity, Alternative Investment Analyst Review, 2(1),

2131.

Barberis, N., 2000, Investing for the Long Run when Returns Are Predictable, Journal

of Finance, 55(1), 225264.

Barberis, N., R. Greenwood, L. Jin, and A. Shleifer, 2014, X-CAPM: An Extrapolative

Capital Asset Pricing Model, Journal of Financial Economics.

Basak, S., 2005, Asset Pricing with Heterogeneous Beliefs, Journal of Banking and Finance, 29, 28492881.

Beber, A., and M. Pagano, 2013, Short-Selling Banks from Around the World: Evidence

from the 200709 Crisis, Journal of Finance, 68(1), 343381.

Beeler, J., and J. Y. Campbell, 2012, The Long-Run Risks Model and Aggregate Asset

Prices: An Empirical Assessment, Critical Finance Review, 1(1), 141182.

Bhamra, H. S., and R. Uppal, 2014, Asset Prices with Heterogeneity in Preferences and

Beliefs, Review of Financial Studies, 27(2), 519580.

Brandt, M. W., A. Goyal, P. Santa-Clara, and J. R. Stroud, 2005, A Simulation Approach to Dynamic Portfolio Choice with an Application to Learning About Return

Predictability, Review of Financial Studies, 18(3), 831873.

Brennan, M. J., 1998, The Role of Learning in Dynamic Portfolio Decisions, European

Finance Review, 1(3), 295306.

Brown, K., and C. Tiu, 2013, The Interaction of Spending Policies, Asset Allocation

Strategies, and Investment Performance at University Endowment Funds, Working

Paper 19517, National Bureau of Economic Research.

54

Buraschi, A., R. Kosowski, and F. Trojani, 2014, When There Is No Place to Hide: Correlation Risk and the Cross-Section of Hedge Fund Returns, Review of Financial Studies,

27(2), 581616.

Buss, A., and B. Dumas, 2014, Equilibrium with Transactions Fees: A Qualitative and

Quantitative Assessment, Working Paper, INSEAD.

Cejnek, G., R. Franz, O. Randl, and N. Stoughton, 2014, A Survey of University Endowment Management Research, Journal of Investment Management.

Christiano, L. J., and J. D. Fisher, 2000, Algorithms for Solving Dynamic Models With

Occasionally Binding Constraints, Journal of Economic Dynamics and Control, 24(8),

11791232.

Cochrane, J. H., 2005, The Risk and Return of Venture Capital, Journal of Financial

Economics, 75(1), 352.

Cochrane, J. H., F. Longstaff, and P. Santa Clara, 2008, Two Trees, Review of Financial

Studies, 21, 347385.

Collett, D., C. Lizieri, and C. Ward, 2003, Timing and the Holding Periods of Institutional

Real Estate, Real Estate Economics, 31(2), 205222.

Collin-Dufresne, P., M. Johannes, and L. A. Lochstoer, 2013, Parameter Learning in General Equilibrium: The Asset Pricing Implications, Working Paper, University of Lausanne and Columbia University.

, 2014, Asset Pricing When This Time is Different, Working Paper, University

of Lausanne and Columbia University.

Cuoco, D., and H. He, 2001, Dynamic Aggregation and Computation of Equilibria in

Finite-Dimensional Economies with Incomplete Financial Markets, Annals of Economics and Finance, 2(2), 265296.

Detemple, J. B., 1986, Asset Pricing in a Production Economy with Incomplete Information, Journal of Finance, 41(2), 383391.

Detemple, J. B., and S. Murthy, 1994, Intertemporal Asset Pricing with Heterogeneous

Beliefs, Journal of Economic Theory, 62, 294320.

Dhar, R., and W. N. Goetzman, William N., 2005, Institutional Perspectives on Real

Estate Investing: The Role of Risk and Uncertainty, Yale ICF Working Paper No.

05-20.

Dothan, M. U., and D. Feldman, 1986, Equilibrium Interest Rates and Multiperiod Bonds

in a Partially Observable Economy, Journal of Finance, 41(2), 369382.

55

Dufwenberg, M., T. Lindqvist, and E. Moore, 2005, Bubbles and Experience: An Experiment, American Economic Review, 95(5), 17311737.

Dumas, B., A. Kurshev, and R. Uppal, 2009, Equilibrium Portfolio Strategies in the Presence of Sentiment Risk and Excess Volatility, Journal of Finance, 64(2), 579629.

Dumas, B., and A. Lyasoff, 2012, Incomplete-Market Equilibria Solved Recursively on an

Event Tree, Journal of Finance, 67(5), 18971941.

Ehling, P., A. Graniero, and C. Heyerdahl-Larsen, 2014, Asset Prices and Portfolio Choice

with Learning from Experience, Working Paper, London Business School.

Epstein, L. G., and S. Zin, 1989, Substitution, Risk Aversion and the Temporal Behavior

of Consumption and Asset Returns: A Theoretical Framework, Econometrica, 57,

937969.

Florance, A. C., N. G. Miller, R. Peng, and J. Spivey, 2010, Slicing, Dicing, and Scoping

the Size of the U.S. Commercial Real Estate Market, Journal of Real Estate Portfolio

Management, 16(2), 101118.

Franzoni, F., E. Nowak, and L. Phalippou, 2012, Private Equity Performance and Liquidity

Risk, Journal of Finance, 67(6), 23412373.

Garleanu, N., and L. H. Pedersen, 2014, Dynamic Portfolio Choice with Frictions, Working Paper.

Gennotte, G., 1986, Optimal Portfolio Choice Under Incomplete Information, Journal of

Finance, 41(3), 733746.

Ghysels, E., A. Plazzi, W. N. Torous, and R. I. Valkanov, 2012, Forecasting Real Estate

Prices, in Handbook of Economic Forecasting, ed. by G. Elliott, and A. G. Timmermann. Elsevier, vol. II.

Goetzman, W. N., and S. Oster, 2012, Competition Among University Endowments,

Working Paper 18173, National Bureau of Economic Research.

Greenwood, R., and S. Nagel, 2009, Inexperienced Investors and Bubbles, Journal of

Financial Economics, 93(2), 239258.

Harris, R. S., T. Jenkinson, and S. N. Kaplan, 2014, Private Equity Performance: What

Do We Know?, Journal of Finance, 69(5), 18511882.

Harrison, J. M., and D. M. Kreps, 1978, Speculative Investor Behavior in a Stock Market

with Heterogeneous Expectations, Quarterly Journal of Economics, 92, 323336.

Haruvy, E., Y. Lahav, and C. N. Noussair, 2007, Traders Expectations in Asset Markets:

Experimental Evidence, American Economic Review, 97(5), 19011920.

56

Heaton, J., and D. J. Lucas, 1996, Evaluating the Effects of Incomplete Markets on Risk

Sharing and Asset Pricing, Journal of Political Economy, 104(3), 443487.

Hege, U., and A. Nuti, 2011, The Private Equity Secondaries Market During the Financial

Crisis and the Valuation Gap, Journal of Private Equity, 14(3), 4254.

Hussam, R. N., D. Porter, and V. L. Smith, 2008, Thar She Blows: Can Bubbles Be

Rekindled with Experienced Subjects?, American Economic Review, 98(3), 92437.

Jarrow, R. A., and A. Rudd, 1983, Option Pricing. Irwin, Homewood, IL.

Kaplan, S. N., and A. Schoar, 2005, Private Equity Performance: Returns, Persistence,

and Capital Flows, Journal of Finance, 60(4), 17911823.

Kaustia, M., and S. Kn

upfer, 2008, Do Investors Overweight Personal Experience? Evidence from IPO Subscriptions, Journal of Finance, 63(6), 26792702.

Kreps, D., and E. Porteus, 1978, Temporal Resolution of Uncertainty and Dynamic Choice

Theory, Econometrica, 46, 185200.

Lerner, J., A. Schoar, and J. Wang, 2008, Secrets of the Academy: The Drivers of University Endowment Success, Journal of Economic Perspectives, 22(3), 20722.

Longstaff, F. A., 2009, Financial Claustrophobia: Asset Pricing in Illiquid Markets, American Economic Review, 99(4), 11191144.

Lucas, R. J., 1978, Asset Prices in an Exchange Economy, Econometrica, 46, 14291445.

Malmendier, U., and S. Nagel, 2011, Depression Babies: Do Macroeconomic Experiences

Affect Risk Taking?, Quarterly Journal of Economics, 126, 373416.

, 2014, Learning from Inflation Experiences, forthcoming in Quarterly Journal of

Economics.

Metrick, A., and A. Yasuda, 2010, The Economics of Private Equity Funds, Review of

Financial Studies, 23(6), 23032341.

Miller, E., 1977, Risk, Uncertainty, and Divergence of Opinion, Journal of Finance, 32,

11511168.

Minka, T. P., 2003, Bayesian Inference, Entropy, and the Multinomial Distribution, Online tutorial.

Morris, S., 1995, The Common Prior Assumption in Economic Theory, Economics and

Philosophy, 11, 227253.

Moskowitz, T. J., and A. Vissing-Jorgensen, 2002, The Returns to Entrepreneurial Investment: A Private Equity Premium Puzzle?, American Economic Review, 92(4),

745778.

57

Phalippou, L., 2009, Beware of Venturing Into Private Equity, Journal of Economic

Perspectives, 23(1), 147166.

Phalippou, L., and O. Gottschalg, 2009, The Performance of Private Equity Funds, Review of Financial Studies, 22(4), 17471776.

Preqin Special Report, 2013a, Private Equity Secondary Market, https://www.preqin.

com.

, 2013b, US Hedge Fund Industry, https://www.preqin.com.

, 2014, Global Private Equity Report, https://www.preqin.com.

Scheinkman, J. A., and W. Xiong, 2003, Overconfidence and Speculative Bubbles, Journal

of Political Economy, 111, 11831219.

Seru, A., T. Shumway, and N. Stoffman, 2010, Learning by Trading, Review of Financial

Studies, 23(2), 705739.

Smith, V. L., G. L. Suchanek, and A. W. Williams, 1988, Bubbles, Crashes, and Endogenous Expectations in Experimental Spot Asset Markets, Econometrica, 56(5),

11191151.

Swensen, D. F., 2009, Pioneering Portfolio Management. The Free Press.

Vayanos, D., 1998, Transaction Costs and Asset Prices: A Dynamic Equilibrium Model,

Review of Financial Studies, 11(1), 158.

Vayanos, D., and J. Wang, 2011, Theories of Liquidity, Foundations and Trends in Finance, 6(4), 221317.

, 2012, Liquidity and Asset Returns Under Asymmetric Information and Imperfect

Competition, Review of Financial Studies, 25(5), 13391365.

, 2013, Market LiquidityTheory and Empirical Evidence, in Handbook of the

Economics of Finance, ed. by G. M. Constantinides, M. Harris, and R. M. Stulz. Elsevier, Amsterdam, vol. 2B, chap. 19, pp. 12891361.

Vissing-Jorgenson, A., 2003, Behavioral Finance: An Impartial Assessment or: Does Irrationality Disappear with Wealth? Evidence from Expectations and Actions, NBER

Macroeconomics Annual.

Weil, P., 1990, Nonexpected Utility in Macroeconomics, Quarterly Journal of Economics,

105, 2942.

Welch, K. T., 2014, Private Equity Diversification Illusion: Economic Comovement and

Fair Value Reporting, Working Paper, Harvard Business School.

Williams, J. T., 1977, Capital Asset Prices with Heterogeneous Beliefs, Journal of Financial Economics, 5(2), 219239.

Zapatero, F., 1998, Effect of Financial Innovations on Market Volatility when Beliefs are

Heterogeneous, Journal of Economic Dynamics and Control, 22, 597626.

58

- ResumeUploaded byPayas Talwar
- Glosssry-IVCAUploaded byapi-3865133
- 2010 Preqin Infrastructure Review Sample PagesUploaded byNoaman
- CEO/Busines LeaderUploaded byapi-78367585
- Top 50 Real Estate PE FirmsUploaded bybzit11
- Nyabwanga et al 2013Uploaded byRobert Nyamao Nyabwanga
- HDFC Short Term Fund Apr11Uploaded byNaveen Surapaneni
- Water InfrastructureUploaded byshopitnow
- liquidity2081.pdfUploaded byUday Sharma
- Directory of Indian Law FirmsUploaded bydeepahire
- Demystifying the Financial SectorUploaded byTanya Singh
- 1205-HKVCA VC Brown PaperUploaded byArvind Baghel
- HEALTH INSURANCE Healthcare Industry Structure_US CongressUploaded byAaronMulvihill
- Camels MelsUploaded byMallu Channi
- LiquidityUploaded bysmysona
- Project Report on NPAUploaded byseepi345
- BF - finalUploaded bySaman Naqvi
- 1864446Uploaded byBiondi Mahesa
- A Demon of Our Own Design Book NotesUploaded byad9292
- Shajalal Islami Bank Ar 2013Uploaded byamriyan123
- MB0045 Financial ManagementUploaded bySolved AssignMents
- 007 pw 2909Uploaded byapi-291251278
- PFRDA Pension Bulletin Sep2015Uploaded bymihirupadhyay9
- A better way to understand internal rate of return.pdfUploaded byfshirani7619
- Flow of the IntroductionUploaded bypaulsd
- Resources_Private+Equity+Essentials_09.19.17_v1Uploaded bymylesx66
- 9-205-058Uploaded byPhương Anh
- PROFITABILITY AND OP EFFICIENCY REVISION.docxUploaded byandeng
- Chapter 7: Answer SheetUploaded bymohamed saed
- WCMUploaded byricha sharma

- Course Work Assessment Task 3 (1)Uploaded byJigyasa Gautam
- IFRS Global Rules Local Use AAU Prague 2015Uploaded bycecilianica
- Bank of America Psychopath Murderer's Automated Email Reply Says He is an _Insane Psychopath_ _ Zero HedgeUploaded bycaldaseletronica280
- EDHEC Position Paper Who Sank the Boat FUploaded byDouglas Ashburn
- Chapter 7 SolutionsUploaded byAndi Luo
- Bloomberg.hedge Fund Risk Fundamentals - Solving the Risk Management and Transparency ChallengeUploaded bymeetwithsanjay
- Section C5 Group1A Case 1Uploaded bysssales
- FIN307 Exam1 (1)Uploaded bymahirahmed51
- Benefits of Dark PoolsUploaded bysouto123
- Financial DerivativesUploaded byMslolita
- IFBL Risk Management 2009Uploaded byorazikma
- Platinum Crypto Prospectus | Comprehensive Cryptocurrency Training ProgrammeUploaded byplatinumcryptoacademy
- Black Swans and Red Ink US 4_15Uploaded bySiddhant
- Objectives of a FirmUploaded byyasheshgaglani
- Damodaran's Optimal Capital StructureUploaded byRajaram Iyengar
- Money Market vs Capital MarketUploaded byraj
- woodie_CCI_2.1Uploaded bydjsinha
- HM - 13-1Uploaded byzvonomir
- Futures and OptionsUploaded byMukul Soni
- Introduction to FinanceUploaded byNuruzzaman Md Sheikh
- International Business Environment 1Uploaded bySammir Malhotra
- Forex, Foreign Exchange, Currency, Technical Analysis, Elliot Wave Theory, Trading GuideUploaded bymohora
- SMC Wealth - Lead... Delicately Poised for Record HighsUploaded byTejas Seth
- Bond Valuation FilledUploaded byFiqri
- Operational Risk FRM L2 Part 1of3 2 NovUploaded byShivgan Joshi
- Chap6Uploaded byjhouvan
- Financial Market EnvironmentUploaded byips4788
- Identifying Text StructureUploaded byTíff Ặŋŷ
- Report - Revobot-2 Optimized SettingsUploaded byAnthony Horne
- Case for Momentum Investing by AQR (Summer 09)Uploaded byrodmorley

## Much more than documents.

Discover everything Scribd has to offer, including books and audiobooks from major publishers.

Cancel anytime.